## So, as of now, Magic is officially dead

Tonight the last Magic World Champion was crowned, as was the last Magic World Team Champion, and also the last Player of the Year. Worlds will be replaced by a boring rehash of the Duelist Invitationals, the Player of the Year will be replaced by something yet unknown even to Wizards OP – or maybe just by nothing at all – and the team competition will cease to exist completely. Make no mistake: This is it, the death of Magic, the mental sport. No matter what exactly it will be replaced with, it will be based on the tenet of rewarding mediocrity in large quantity and marginalizing excellence.

You are no longer a student of Magic strategy on your road to Worlds.

You are a planeswalker.

**Explore posts in the same categories:**General

**Tags:** excellence, Magic, mediocrity, mental sport, Planeswalker, Worlds

November 21, 2011 at 6:09 pm

You’re right, sad day. A new chapter in the history of the game will begin. I feel like it’s goning to be like when Star Wars Episode I was created. It has the same name, it is somehow similar, but on the other hand completely different, with many important pieces that contributed to what made it special missing.

I don’t want to be a planeswalker

ET

November 22, 2011 at 7:40 am

Yeah it is true the last worlds, as we know it, are done.

But i have nothing against an invitational Worlds (Not many “Sports” have 150+ Players in there Worlds). PotY and RotY are History, but lik you said there will be something new, so lets wait for it, before we call the Game Dead Again. MAybe its all harder then we thought.

Maybe the Only thing we can sad about is the Meaningless NEW National Championchips.

November 22, 2011 at 12:18 pm

The thing is, magic was never a mental sport. It was always a pray on the weak. The only mental threshold magic has, is to understand the rules and mechanics. After this the game is a coin flip. Lock at the data from worlds all decks with sufficient numbers of matches played, have a win percentage in the range of 50%. And even the best deck played by the best players have 60%! So your margin for being skilled and prepared is 10%. That’s a joke of a mental sport. You can argue that an individual player might be far better than the average player but you will never now this because to determine the real skill of an individual at this tournament you would need a high number (140-220) of matches played for each player to neglect the influences of the intrinsic variance of the game, match ups, starting player and so on. This is an impossible number for any tournament. But over a long time it will pay of and you get your 10% more. And that is exactly the margin the top players have. But this margin will not win you single games. For a mental sport you need one fundamental truth: If you don’t make a mistake you win or at leased you will not loose. Magic can’t give this promise. So Magic was always a luck based game with a limited skill offset. That’s enough to play it when you want to relax with your friends and have a nice evening but it is never, never, never enough to be a mental sport.

November 22, 2011 at 12:36 pm

Just like poker, I suppose? There’s no doubt that in poker skill is an even more incremental advantage, and still there are pros in this game being able to make a living of it.

A mirrormatch (single game) between a player who just understands the rules and a superb player can be anywhere between 50/50 (totally non-interactive decks) to 75/25 (extremely intercative and skilltesting decks).

An overall win percentage of 60% will have you pull ahead of the field very distinctively over the course of even a few tournaments. Elo rating, despite its minor flaws, was able to represent this.

The whole point of the new system is to hide those skill differences, because they are still way too apparent and bad players find it frustrating to acknowledge that others are better than them.

Last year in the Bundesliga, the difference between a 60% win percentage and a 40% win percentage came down to roughly the difference between 5th place and 14th place. That doesn’t mean there is little skill in soccer, or that it wasn’t a real sport, even though Mainz got a little lucky last year and Schalke a little unlucky. The skill differences in Magic are apparent enough to allow for ranking players according to their skill.

November 22, 2011 at 1:24 pm

But poker allows far more games in the same time. So its even more easy to extract the good players. But thats not the point the point is that these pros have to grind hard and have to constantly reinvest their winnings. Only if they can play a huge number of tournaments their skill will mean anything. So it alway comes down to grinding like a mad man. So the real skill of these players is not the mental part but the endurance and willingness to attend this high number of tournaments.

The old ELO rating had also many flaws. The biggest point was that it had no uncertainty measurement of your rating. In a game with high variance you are never sure if a winning or losing streak is the cause of chance or skill. So you have to factor this in. A rating with 2000 and 95% certainty is a good measurement a rating with 2000 and 50% certainty is useless and has to be proven by further games. But this would only lead to a similiar system like the planeswalker points and you would be forced to play a huge number of games in every season to get your certainty up.

Both systems the new and the old one are far from a fair and scientific measurement of skill. Both are equally worse only in different parts.

But than is still the question how fair is a mental sport if you can’t prove your skill in a single game or even a single tournament.

November 22, 2011 at 2:28 pm

How fair is a soccer game if it is possible that an amateur team beats a champions legaue team? How fair is the DFB-Pokal?

The old Elo system used in Magic did have flaws, but those could have been addressed in a variety of ways. Since that was never the real issue, though, it was just scrapped, its flaws just used as an excuse to install the new system.

See, you don’t have to play THAT many matches to make it possible to assess your skill (of course there is an amount of uncertainty, but there always is, in every sport, with every system). You know what does the trick? Right – the Elo system! That’s what it was made for, and if adjusted correctly, it works fine.

There’s a vast difference between “grinding like a mad man” and playing in enough tournaments that you can expect to show your skill. You know which system allowed Magic pros to do exactly that? The Pro players’ club! While playing a reasonable amount of high calibre tournaments a year, fitting into a pro’s RL schedule (which could mean just the PTs and a random GP here and there, or travelling the worlds for GP almost every weekend), they could amass enough points through consistent finishes to be able to keep up being a pro. Of course, playing in more tournaments gave you an edge, but an excellent player could still prove his excellence without grinding and be rewarded.

All this has been changed now, so that grinding like a madman has become the ONLY way. You don’t seem to understand: The old system was razed because it – by and large – WORKED! Skill differences between players were clearly apparent. Top players could stay on the gravy train AND have a real life.

That’s what they wanted to change. Bad players should no longer see that they were bad. Success should only come to those playing Magic 24/7.

The changes are anything but a confession that Magic, the mental sport, didn’t work; they are the consequence of WotC deciding that it should no longer exist!

November 22, 2011 at 3:45 pm

The old ELO system was not a measurement of skill and the old Pro players system was also not a system to prove that you are a got player but only a grinder. The old pro player system had a very strong glass ceiling problem. If you where in the system it gave you more EV by attending your invite PTs as for a typical PTQ player. And that was completely unrelated to your rating. A 1900 player could be on the train and a 2000 player who just regularly makes Top8 but never wins wouldn’t even be on 1 PT.

I have written my PhD thesis based on a lot of statistical mathematics and can assure you with 100% certainty that the old ELO system does not work in a game with variance. It isn’t even near of being accurate. Even in chess the ELO does not work and extensive studies are being made to find a better system. It works only in the theoretical case of very high numbers played and very small k-values. To get a reasonable extrapolation of your real skill with ELO in magic you would need a sample size of over 40000 games and k-values around 4. And this factors only the variance of screw/flood games, starting player and match ups. You can easily find more factors which influence the outcome of a magic game beside the skill. And which each factor the necessary number of games increases exponentially.

In the old systems Brain Kibler and other Pros traveled around the world to farm pro points at GP and PTs to get their level in the next season. Now they travel around and get PW points. The only difference is that before you could get a high level by getting lucky at one tournament and that people who could not travel as much had never a chance of being on the train. Try to calculate with a reasonable EV in both systems and you will see that the number of grinding doesn’t change that much. And wizards have clearly stated that if they see cases of PW points abuse they will change it accordingly so that only PTQs and above feed into PTs with a reasonable amount of play.

Yes I know their is the case of the FNM and sideevent grinder but this way will cost you thousands of euro if you only grind in this way and are a bad player. So in principle its like you have to pay the full buy-in in a poker tournament and others get it paid because they won a qualifier. No one has problems with that.

Magic has to change more towards a poker like system and wizards have realized this.

November 22, 2011 at 4:21 pm

How can you edit your own post?

The last sentence in the first paragraph should be. A 1900 PT player could be on the train and a 2000 PTQ player who regularly makes Top8 but never wins wouldn’t even be on 1 PT.

November 22, 2011 at 9:36 pm

Where have you read that “if they see cases of PW points abuse they will change it accordingly so that only PTQs and above feed into PTs”?

Wizards talked vaguely about some kind of adjustments, but this is the first I hear about any meaningful positive change.

November 22, 2011 at 11:17 pm

Q: The 3x multiplier for Friday Night Magic is confusing some people. Does it mean that we are forcing pros (and aspiring pros) to play in FNM events in order to qualify for the Pro Tour?

HB: Let me make it very clear here that FNM is not how you get onto the Pro Tour. We don’t expect pro players to have to attend FNM to continue to participate in the Pro Tour, and grinding FNMs should not get you to the Pro Tour. Friday Night Magic will remain a fun and friendly non-pro-level competition, and if it turns out that the Planeswalker Points system as it exists today doesn’t help create this environment, we’ll make changes.

directly from: http://www.wizards.com/magic/magazine/article.aspx?x=mtg/daily/feature/167b

November 23, 2011 at 5:25 am

It really depends on what your idea of a “reasonable” approximation is. As you say, there’s variance even in chess. In most sports, people are satisfied with a rough approximation of skill. Why do you ignore that? Soccer is an excellent example – there’s a lot of variance in each game (also, a team’s stength isn’t constant, which is also the case with a chess or Magic palyer’s skill), but still the standings in a league give a good enough approximation for practical purposes. Of course, no accurtae system can even exist, just because players’ skills aren’t constant (and also, because relative skill levels between players aren’t transitive, even disregarind different decks). Variance just adds to that issue.

Still, a clearly higher Elo value corresponds with a high propability with better play skill, even though the exact model used for Magic does a bad job of reflecting this game’s idiosyncrasies, just as a clearly better standing of a soccer team correpsonds with high probability with a team’s strength. In short: Elo works reasonably well (with room for improvements).

The glass ceiling problem is an issue which could have been addressed, but I don’t see anything wrong with it in general. Why shouldn’t the standing of a level-x pro be something attainable by success? It is just like e team in a lower league qualifying for a higher one. It might be unfair to other temas as good as – maybe even better – than the one actually qualifying, but there is a certain ranking which allows for elevation to a higher (and more lucrative) play level. Becoming level-x pro is that qualification (and it is NOT totally unrelated to you rating, since rating gives byes, and byes really help in doing well in events).

People much better suited for this already have made calculations (did you read Zvi’s article?). Anyway, the point is not that the most succesful player has to grind in both system. The point is that an excellent player, who can only play a reasonable number of tournaments, had a good shot at keeping a pro level high enough that playing tournaments all around the world wasn’t EV- to him, but under the new system, ONLY grinders can do well, driving top players with a real life out of the game.

Also, the ONLY thing that Wizards have clearly stated so far is that there would be changes, but they can’t say which so far, becaus ethey do not even KNOW. They claim they realized they made a mistake (although, in my opinion, they just tested the waters to see how big the backlash would be to act accordingly afterwards).

Oh, and you obviously cannot edit your comment, only I can. Remember that, in the end, I am responsible for everything posted on my blog, and you should understand that, even if there was a way to allow posters to edit their comments (there isn’t) I wouldn’t use that option!

November 23, 2011 at 12:43 pm

A reasonable ELO system can predict the outcome of a tournament with little error margin. In this case the game is won by skill and the ELO reflects the true skill value of the player compared to other players. In magic this is not true because it has a too high variance. And this is also the reason why an ELO like system is not used most team sports but they are regularly used in single player competitions like golf or tennis. Just to give an impression of how wrong the ELO system is. Given a completely gauss distribution of 1 Mio fictional players each with a true rating but starting by 1600 ELO. Thereby we assume that the players have a skill offset accordingly to their true rating. So no learning by playing or bad days or local phenomena are considered. Now we let them play in a set of random FNM, PTQ and GP tournaments. After 1000 games each, 50% of the players have a rating which is on average 300 points above or below their true rating. After 10000 games there are still 50% wrong by 190 points. After 100000 games it slowly goes down to 150 points. The bad thing is the ELO can’t be fixed to be a better measurement system for magic. It was already a devils compromise between get the players fast to their true rating and be an “accurate” measurement of true skill. With the result that each player with 1800 rating is to 50% wrong by 200 ELO points and each player around 2100 is with 37% overrated by 150 points. How can anyone justify giving rating invites or buys based on such a system.

To get the discussion back on track and not get lost at side wars.

Your arguments are, and correct me if I am wrong.

1.) Magic is a mental sport

2.) ELO ist the better system to measure skill compared to the PW system

My arguments are:

1.) Magic is gambling with a small skill offset of 10 to 15% at maximum which can be proved by live time win percentage of very good players, and further backed up by group performance of pro players at big tournaments (to note group performance is not a hard evidence just a weak indicator). The small skill offset compared to its intrinsic variance disqualifies magic as a mental sport because it is not possible to prove your skill with the small sample size generates by usual magic tournaments.

2.) Skill in magic can never be measured in a practical way by any method. It would require such a huge dataset that it would always make insane grinding a necessity. You are right that ELO can be better than PWP in magic but it is the same as saying 1+1=3 is more right than 1+1=4.

My opinion is that organized play should move away from PT, Nationals, Worlds and any rating at all and just establish a bigger GP circuit with slightly higher prize money, and with GPQ where you can only win the travel expanses to the GP. Which are preferential given to shops but can also be held by private organizers, if the shops are unable/unwilling or just not in a certain vicinity. In this system everyone gets the same starting chance for the “big money”. You can improve your chances by grinding GPQs. But this grinding is not a necessity. The Hall of Fame can than also be remodeled and given to players which reached a certain live time win percentage, top8 appearance or money winning at GPs. All under the assumption that they played a certain amount of matches at GPs. This system is much more fair for the whole player base and very simple, but it is still magic as a competitive game at its maximum level. I can only assume that wizards is moving with the PWP system towards such a model.

November 23, 2011 at 5:11 pm

I’m pretty sure you can assume Wizards haven’t, in any way, have entertained similar ideas than you express in your post (although they may jump on it if they are confronted with it).

I admit there is a point where I can not follow your math (partly due to the fact that you give no explanation for your claims, but it may well be that I couldn’t follow it). However, the results you get seem really fishy to me!

You say 50% of all players worldwide would have a rating wrong by 300 points or more, after 1000 matches ( I assume you mena matches, not games) each, even if assuming playing skill was a coonstant? Can you somehow back that claim up? This is an error margin which isn’t just, as yu say, unacceptable; it is downright unbelievable. I’ve seen people who knew a lot more about a certain area than me been wrong before time and time again since they never felt the need to validate their claims, and thus I hold it possible that you might be working under false assumptions.

Elo is, after all, a self-correcting system, and the issue with it isn’t that it OVERSTATES skill differences, but UNDERSTATES them. I know that a rating difference of 200 points ( a few points less in Magic, because they use a slightly different formula than chess) is calibrated to define a 75% win percentage of the better rated player. Thus, if actual win percentages in general have to be lower (which we both agree they are, due to the fact that there is a lot of randomness in Magic), the system would work AGAINST large skill differences bewteen players, and I really don’t see how half of the whole playerbase could be assessed wrongly by that amount!

The issue with Elo ratings and Magic is, that with random pairings between players, a difference of 300+ points is near unattainable and unsustainable, for the very effect I just described! The existing more extreme differences are due to two reasons:

1. Players with higher skill and higher ratings preferredly, and often exclusively, playing againsgt each other in tournaments with very high k-value. (This is why there is an effective “ceiling” for a player’s rating who is not on the gravy train of 1900-sth, which is a phenomenon that could actually be observed.) This is also why extremely high ratings (anything clearly above 2000) fluctuate wildly. HERE you have wrongly attributd skill differences, but for tweaks in the tournament system your model doesn’t reflect, and of course for a much, much lower number of players!

2. People leaving the system by stopping to play, leading to an increase of the ratings average (because, on average, many more unskilled players stop playing than skilled, effectively pumping points into the system during the time they play).

So please give an explanation of how you arrived at your astonishing conclusions with your model, and do not forget to list all your assumptions! Because what you claim, really, REALLY looks like bullshit to me.

November 23, 2011 at 5:24 pm

Oh, and by the way, if your explanation holds any water, you might want to consider adding it to the Wikipedia article about the Elo rating system, because your finds seem to be unknown so far by that community… It seems that, in the wide world, people actually entertain the idea that the Elo rating system works well and largely as intended!

Also: roughly 82% of all rated players in the DCI have a rating inbetween 1500 and 1700. This means that at least 32% off all players which, according to you, have a rating off by 300 points or more, fall into this range (or would, if they just played enough games). I’m really interested in your explanation!

November 23, 2011 at 9:22 pm

The thing is easy. ELO is only used in chess like sport where the k values getting smaller the more pro you go. That is the real ELO system. In Magic you have huge k-values swings at pro tournaments which makes the whole system unstable. But that is not the only reason. The other part is the whole variance issue. I think you are familiar with the law of big numbers. You need an ungodly amount of matches (and yes I mixed matches with games in the last post) to overcome this variance issue. In chess you don’t have this issue because you can safely assume that if a player beats another its because of his skill.

For a small sample size (< 1000) the ELO system is just chaotic (for magic). I know for normal people 1000 seams huge but is just a drop in the ocean.

I haven't actually calculated this results I have simply simulated the results. And yes I can only say from these results that 50% of all ratings in magic are wrong by 200-300 points. Simply from the facts that they didn't played enough matches or had recently a huge ELO value swing.

November 23, 2011 at 9:44 pm

There isn’t “a real” Elo system; there are several different implementations used in a variety of sports, and even several different versions used in chess. Also, k-values are not “getting smaller” in chess; there are just two different k values for players according to their rating. I identified the issue of too high k-values in big Magic tournaments above, however, this obviously affects just a very small minority of all players.

As for your simulation: Sorry, but the credibility of its outcome depends on your exact method and of your parameters, and from this discussion and comparing your results with Elo behaviour in the real world I have no choice but to assume that you got those fundamentally wrong!

Also, you identify yourself too high k-values as the main cause of huge Elo swings (although you fail to realize that this issue only affecst a very small percentage of players, not at all explaining your findings) – but that isn’t a flaw of the system overall, but of its implementation for Magic.

I’m absolutely convinced that in a closed system of players with a minimum of 1000 matches each, variance will not cause nearly as bad swings as you claim for a relevant number of players. Unless you lay down the exact method of your simulation, your claims are just absurd!

November 24, 2011 at 12:57 am

Just consider this: In magic you are not sure that the rating of your opponent is correct, that is an uncertainty even chess has and you are not sure that the rating he gained in the last time was due to skill or chance the same is true for your own skill and than you are not sure that the outcome of the match you played against this opponent was skill or variance. That are 5 uncertainties in one measurement. It gets even worse all the opponents you faced until this point and the opponents you faced have the same problem. So your ELO depends greatly on the string of opponents you face and what ELO they have. Normaly the ELO would correctly predict the win percentage of each match and each player would loose rating accordingly but in magic the variance of the game has a greater influence as skill. So each single measurement is flawed and the system can’t correct itself because its based on the axiom that skill is the only variable that determines the outcome. And on top of that you have huge swings due to different k-values.

The simulation just showed this huge chaos factor. And unless you give me a mathematical proof that it is possible to define a measurement point between 2200 and 1200 with an error margin of 10% (because that is the value when fair begins) when 5 independent variables are influencing greatly your measurement and using only 1000 data points, which even depend on millions of uncertain data points, you are absurd.

I studied this kind of thing, I wrote my PhD thesis using greatly such mathematics and I’m using this kind of mathematics every day at work. Of course I cant be wrong. But you have to argue with me on a scientific and academic level and not use some cheap trick. So you are absolutely convinced that 1000 points are enough so the rating you have should by correct. What is the mathematical method that gives you that kind of certainty. And if this is true. Given a player with a rating of 400 points more. Are you convinced that you would loose against him in a mirror match in 1 of 100 matches. Because that is the win percentage the ELO rating gives you. No you play most likely 40:60 or better against him. Lifetime win percentage of pros shows clearly that it doesn’t get better. But than your ELO was WRONG and the ELO of your opponent was WRONG. But you were convinced your rating was correct because you had 1000 matches played already and represented you real skill value against all the other players worldwide. So how could it be that you had another win percentage as predicted. Try to disprove just this small problem with hard arguments.

November 24, 2011 at 2:07 am

Okay:

1. You are babbling. You’re not following any clear line of reasoning. And you refuse to explain how your simulation is set up.

2. That you make your money at something does, in no way, that you cannot be biased or simply overlook something. I have a lot of experience with experts whose trust in their competence has made them unable to see obvious flws in their assumptions, who neglect common sense, and who flat out refuse reality, because it doesn’t fit into their paradigma, untils forced to admit it by being presented with undeniable facts. Actually, most of these were programmers.

3. Elo system has been developed, and is being put to use, by people with even more experience with these things than you.

4. You show clear signs of simply not understanding the assumptions of the Elo system (explaining why your simulation produces completely unbelievable results). It dies not make any assumptions about skill at all; it is based on the assumption that players have certain win probabilities against each others. It does not try to explain where these win probabilities come from, and its self-correcting mechanism make sure that the win probabilities it uses do NOT diverge to much from a player’s actual win probability. Elo does NOTask for an input of a skill level; its only input are actual results, and the win probability it uses (expressed in a rating) is purely based on these results. All your babbling about uncertainties and if a game’s outcome was influenced by skill or randomness (as if it wasn’t both nearly all the time!) has absolutely nothing to do with the quality of the Elo system’s approximation of your win probability.

5. That said, the win probabilities of the Elo system do NOT fluctuate randomly. There is a certain variance, due to a number of factors, and especially on the pro level, due to higher k-values, but the ratings over the last 15 years or so show clearly that win probabilities in a certain margin are actually correlated with single players, proving that skill is an important factor in determining actual win probabilities.

6. Actually, your model’s assumption that a player’s skill is constant does most likely introduce a large systematic error into your model, because huge swings in a player’s rating, especially below pro level, can very well be the result of a change in a player’s skill (due to a learning curve, getting out of toch, doing a different amount of preparation or playing different formats where the palyer is differentöy skilled in).

7. Since Elo represents a player’s actual win probability, the only indication you can really have that a player’s rating does not match his real win probability is wildly fluctuating ratings. You claim ratings of 50% players to be off by 300 points or more. The fact alone that in real life these fluctuations are nowhere to be observed in any relevant numbers, discredits your simulation already.

8. Also, while I’m not really an expert for probabilities, the claim that a system due to “chaotic” factors will lead to such extreme behaviour in over half of all cases is obviously absurd, esepcially in a self-correcting system. The very distribution you claim to use will prevent this with absolute certainty. If anything, chaos would bring all ratings together even closer together, into a margin of not much more than the highest k-value. If your claim was true (and I’m more and more convinced that you just made up your numbers), a full quarter of all players would experience an aberrant upsurge over the course of several dozen games at least (actually, more of a three-digit number) to get a rating higher than their “normal” win probability by an amount of the maximum possible value they can gain in an 48-k tournament times sixpointsomething – and another quarter has to have an aberrant downsurge of the same magnitude! USE YOUR FUCKING COMMON SENSE! You obviously have a systematic error in your simulation (if it exists at all).

9. Get rid of the idea that there is a systemati differenc between Magic and chess when it comes to express a player’s skill by Elo rating, unless you are looking at cases where players avoid playing other players with much lower skill (once again, that is the thing which inflates Magic ratings – oh, and alos chess ratings, by the way, but in Magic, that effect is magnified by a more or less constant amount of games which are decided by sheer luck). A win probability is a win probability, no matter if the game in question is mostly skill (I hope you understand that there is luck in chess, otherwise you really shoudn’t be discussing this topic!) or mostly luck. If Magic was a pure game of luck, all win probabilities would be 50& (and the Elo system would reflect that near perfectly – another strike against your reasoning, because it is the existing skilldifferences which allow for minor distortions at all!) Since it isn’t, win porbabilities range somwhere in between 25% and 75%, where in chess they would range in between, say, 1% and 99%, That does not make Elo not applicable for Magic at all, it just means that ratings will be much mor closer together, if using the same scale.

10. The fundamental problem with the idea of attributing fixed win probabilities to players is that for this system to work perfectly, there had to be a certain transitivity to it: If between A & B, a skill difference of 200 points means a win probability of 75% for A, and a difference of 400 points menas one of 90% (the way the original Elo system was calibrated), then, if A’s rating is 200 points higher than that of B, and that of B is 200 points higher than that C, A should have 75% against B, B 75% against C, and A 90% against C – but in real live, it doesn’t work that way. It would not even work in a hypothetical game where luck is not a factor, it works even less in chess, and it works not at all in Magic, where any player with a fundamental grasp of the rules has something like a base chance to beat every opponent due to screw etc… However, note that this effect a) relies of actual skill differences existing and being reflected by the Elo system, and b) actually works AGAINST the effect you describe, by bringing A’s rating down to an amount which is somewehere in between reflecting his win percentage against B and his win percentage against C.

You are the one who made an outrageous claim and refused to give any explanation to it. Either you finally explain in detail your simulation, or I will take this as a concession that you simply wrote nonsense.

November 26, 2011 at 12:49 am

This discussion can now officially end, since trischai has in a German blog entry explained his simulation, showing that he executed it with the explicit premise that Elo doesn’t work (and thus, not really surprisingly, getting the result that it doesn’t work).

Link to his blog entry: http://magicblogs.de/multiplayerblog/2011/11/25/warum-das-elosystem-fur-magic-nicht-funktioniert/

November 26, 2011 at 12:17 am

done!

http://magicblogs.de/multiplayerblog/2011/11/25/warum-das-elosystem-fur-magic-nicht-funktioniert/

November 26, 2011 at 2:59 am

For readers not able to understand German, a short summary of that article:

1. With chess, Elo works great.

2. With Magic, it would work exactly the same, but since the author wants to prove otherwise, he introduces an additional premise into the system: Namely, that there was some kind of “true” win percentage of a player, which is explicitly NOT identical to his actual win percentage!

3. Since the Elo system, obviously, only processes actual win percentages of players, the ratings it prdouces do not match the “true” win percentages whimsically defined by the author.

4, The author presents the fact that the Elo system is unable to match “true” win percentages (under the premise that they are different from actual win percentages), citing this as a proof that Elo doesn’t worh with Magic.

5. Of course, all he did prove is that Elo doesn’t work if you explicitly expect it to produce arbitrarily predefined results not based on actual win percentages.