Chess as a fantasy sport

The skill lies in predicting the difference at the end of the period from the rating at the beginning.

Essentially, the person who spots the likely omprovers the best should be the one that wins.

The only debate should be the formula, which should essentially not rely on the start rating.

I couls propose a formula off the yop of my head, but it really requires more thought than a gut feel solution.

If you insist I will prpose a formula after due consideration, but I fell sure that a really good Mathematician would come up with the best idea.

I am good with figures, but not an academic mathematician, so that is why I am reluctant to be first cab aff the rank.. 

That's the whole point?

My objection to that is that it is harder to improve at higher levels; thus, actual rating should be taken into consideration. Moreover, if rating difference is the only thing that matters, most of the players chosen would be close to the bottom of the pool of eligible players - essentially names that most people haven't heard of. From a fantasy sports perspective, it is much more fun to have the big names on your team. And for chess fans, it is more exciting to predict the big improvers at the top than those in the 2600+ range. So I think that while actual rating can be deemphasized, it should not be ignored entirely.

Carelsen then, if you "get" your way, is a points winner from the POV of his being the best player in the world today, about 2800, (I have only ball park figures without consulting the very current ratings), bu t he is not a good point scorer from the POV of points gained, which mught even turn out negative.

But a young Chinese up and coming could be the ideal choice.

So current rating and the multiplier for points gained would have to be carefully worked out.

OK, here is my reluctant first try. Improve on it if you will or discard it as you please.

I pick Carlsen as one choice. Say his rating at the start is 2800. I get the impression that it may tumble  by 50 points to 2750. He actually goes down slightly only to 2790. My predicted margin was a fall of 50. The fall is only 10. So I get 40 as my "absolute" error.

I pick Joe Blow start rating 2600. I see him as an impover to 2680. He achieves 2700. My "absolute" error is 20.

Make the Multiplier say 10

Carlsen gives me a score of 2800 minus 10*40 (400) which is 2400.

Joe gives me a score of 2400 minus 10*20 (200) which is 2400.

So I am inclined to think of 10 as being a ballpark mutiplier.

What do you think.

Hope you can follow all of that!

Best!

camter, or a close variation, as i am known in certain arcane mathematical sites, namely OEIS and Patrick De Geest's site. I will not gine my actual name, but camter is a coded name based on an historic typo, at which I in my myopia, I am prone to.

Well, then you're predicting something completely different. You're not predicting who will improve the most, but rather how the ratings of some specific individuals will change. If that's the case, then there wouldn't be a need to consider initial (or final) rating, since predicting Carlsen's rating change isn't inherently more difficult than predicting the 2600's (it's probably easier).

Macer, you have got my idea. Is it of any help? I must say that I would think that Carlsen is unlikely to change by much, bit Joe Blow's rate of inprovement is far harder, much more risk for the predictor. Thay is more exciting if that is what you are looking for.

But, I am not going to be "hurt" if you discard my idea. I am only here to "help".

Your humbug savant, @camter.

I think the best way to do it would be the following:

- Set up a database of last years scores

- Try out a bunch of different formulas (using some kind of program, I could do it in python if you deliver the data)

- Compare the spread of each formula

- A formula that delivers an exciting competition would be the best, that would be one that:

        - would have a lot of variance in the top scoring players (so players from all ratings should be present in the top 10% scoring)

         - would have a standard deviation of about 100 points so that choices really matter

- Also, should competition rankings be taken into account?

Macer, this poster has all the hallmarks of a Professional Mathematician. 

Fractals. Number Theory. Recursive Function. Statistics.

Just the bloke you need.

Hire him!

Believe me, he has what is needed. I do not have it, but I can recognise someone who does!

Oh! The second page is all the more confusing than the first.

ikr

Ahh yes, this thread. I had forgotten about it. It seems that there isn't broad interest in the project, so I'm going to table it for now. But if I do decide to try to launch it again at a later date...

@ Nathan: Do you know how I would be able to do all of the things that you mentioned? Would you be willing to do them yourself? Or do you know someone who might be able to help? Thanks.

I know enough programming to do it. However, I'm not going to scavenge the internet for databases. Also, I'm currently kinda busy so it depends on when you need those answers if I can deliver.