undefined | Better HN

0 pointsgizmo6865y ago0 comments

The game itself has not changed, so it still makes sense to compare players across time. It would be nice if we had a quantitative way of doing this; so we can make statements like 'the average proffessional player today is better than 20 years ago, a typical modern pro would win 60% of the time again one from 20 years ago).

In some sense, it is not surprising that we do not have a system that accomplishes this. Since it is impossible to see the results of a game between players living in different time periods, we cannot get any data to prevent drift. You can still try to normalize the rankings. However, unless you have some independent way of measuring skill, you would need to make an assumption about the relative strength of players. Assuming the average skill of a proffesional is constant across time is probably not accurate, but closer to reality than what you get with unchecked inflation.

0 comments

7 comments · 3 top-level

smabie5y ago· 3 in thread

You can sort of solve the inflation problem by zscoring the elo. Now a person's score will tell you how much better or worse they are than the median player, assuming an underlying normal distribution (reasonable).

Of course, scores will only be comparable if the average skill of all players remain constant. I would imagine this isn't true, but the drift over several decades is probably small.

Unless you start introducing some purely objective criteria for skill, which can never work, this is the best you can do. It's still way way better than a straight elo system though.

wdevanny5y ago

Rating distributions are often not normal because some subset of players study the game and take it more seriously resulting in a bimodal distribution. See [0] for an example in Chess.

[0] https://chess.stackexchange.com/questions/2550/whats-the-ave...

thaumasiotes5y ago

Even without the bimodality, you wouldn't expect a normal distribution of ratings.

1. Assume that chess ability is normally distributed in the population.

2. Assume that people who are terrible at chess are more likely to stop playing chess than people who are successful.

Then you've sampled the underlying normal distribution mostly from the top end, and that new, highly skewed distribution is what you'll see when you measure everyone's rating.

smabie5y ago

That's fascinating, thanks! It looks like you can model it as a mixture distribution made up of two underlying normal distributions.

thetinguy5y ago· 1 in thread

The idea that chess has not changed in a long time is simply not true. Two huge and relatively recent changes were the addition of chess clocks and premoves.

danielbarla5y ago

And aside from the mechanics of how the game is played, there have been massive changes in the popularity of chess (first massively upwards, recently possibly down slightly), as well as how analyses are done.

It would be very difficult to account for these factors in a way that keeps comparisons across 30-year+ time spans meaningful.

ouid5y ago

the game itself has changed quite a bit, and the number of people playing it, and the dominance that they achieve has also gone up quite a bit.

j / k navigate · click thread line to collapse