Author: Roy Eassa
Date: 07:48:09 03/30/02
Go up one level in this thread
On March 30, 2002 at 00:52:23, Uri Blass wrote: >On March 29, 2002 at 19:21:00, José Carlos wrote: > >>On March 29, 2002 at 12:41:09, Uri Blass wrote: >> >>>On March 29, 2002 at 12:13:41, Roy Eassa wrote: >>> >>>> >>>>A theory on this topic: >>>> >>>>The GM should get a small fixed stipend just for playing, plus $0 for each loss, >>>>$x for each draw as White (where x is very small and perhaps 0), $y for each >>>>draw as Black (where y is still pretty small, but larger than x), and $z for >>>>each win (where z is much greater than y). Finally, a BIG bonus for winning the >>>>match overall seems to make sense. >>>> >>>>That would seem to be real incentive to play his (or her) best every game. >>>> >>>>(Feel free to substitute pounds, Euros, whatever!) >>>> >>>> >>>>Another (better?) approach is to reward only the final score. Besides a small >>>>fee for showing up and playing, a GM score of 25% or less is no money, 100% is >>>>BIG money, and the payments are distributed logically between these two extremes >>>>(but the payoff for doing well should increase faster than linearly, IMHO). >>> >>>If the target is to encourage the GM to do the best then I think that payoff >>>that increase linearly is a good idea. >>> >>>If the payoff increase faster than linearly then the GM may take unnecessary >>>risks. >>> >>>The GM may prefer a strategy that give her(him) 30% chances to win and 40% >>>chances to lose in the game and not 10% chances to win and 10% chances to lose >>>that is more logical. >>> >>>Uri >> >> I disagree, because the same that you (correctly) apply to winning, applies >>also to losing. I'll explain: let's suppose that the money increases >>superlinearly with points. Then, for: >> >>0, 0.5, 1, 1.5, 2 >> >>points the money is: >> >>0, 50, 150, 300, 500. >> >> Now, the GM has 1 point. With a loss he gets 150, 300 for draw and 500 for >>win. He sees he can easily draw. Your point is that he'll take innecessary risk >>to get the 500 but he can also think "if I win the game I get 200 more but if I >>lose I get 150 less". So he'll probably pick the draw if he thinks going for the >>win is too risky. > >You may be right in that case because people hate risks and it is not clear if >for them 150$ and 500$ with 50% probability is better than safe 300$. > >I still think that their gain(not in money) should be proportional to their >result in order to convince them to do their best. > >Uri Payoff should be proportional to the (GM's perceived) DIFFICULTY of the result! There's a lot of evidence that getting a draw against a top program is a LOT easier (five times as easy?) for a GM than getting a win. But it scores a full HALF as many points. So paying proportional to the SCORE rather than the DIFFICULTY means the relative payoff for a draw is, currently, TOO HIGH. Analogy: When I was a student at MIT I discovered the power of partial credit. Some homework assignments could take dozens and dozens of hours to finish completely. But they gave partial credit for partial answers, and I discovered that in maybe 2 hours you could get 50% credit -- and you could then use the saved time for other activities, including homework in those courses that did NOT give partial credit. It's also similar to the 80-20 rule. 80% of the result comes from 20% of the time/effort. To get the full result is often MANY TIMES more difficult and time consuming than to get to the "adequate" level. Perhaps in chess it would be the 50-10 rule: to get a score (and, if poorly set up, a dollar compensation) of 50% requires only 10% as much effort as getting a score of 100% in a given game. IMHO.
This page took 0 seconds to execute
Last modified: Thu, 15 Apr 21 08:11:13 -0700
Current Computer Chess Club Forums at Talkchess. This site by Sean Mintz.