[Fluffy]

I am going to try and make a handicap table & ranking for my local area.

To do so I will gather all results from local clubs withing a year or maybe more.

The basic idea is to calculate an avergage % for each player first, then do the same but apply a corrector based on the average of the field in wich he played each tournament*, and hopefully repeat this procces untill it conveges to something giving me an average for each player.

The question I want to ask is this:

If each field had equal level but then I make a special tournament with mostly good players, is it aproximatelly true that a player with an average 50% put on a field of average 54% would be "expected" to make 46%?


* I might put something based on the levle of partner as well, but that will come later.

[ArtK78]

That is a resonable expectation. In any event, for purposes of handicapping the field, it is a reasonable position to take.

However, you should not use a full handicap. In your example, it appears that you are giving the 50% pair a full 4% handicap. You should only give about 80% of the handicap that a pair merits so that the handicap does not become insurmountable. So, in your example, the 50% pair would have a handicap of 3.6%, not 4%.

Even 80% may be high.

My unit routinely runs handicapped KOs in our local sectional tournaments. The handicaps are based on masterpoint holdings. I don't know exactly how the handicaps are calculated, but I have been in a position of having to overcome 15 IMP handicaps over 24 boards. Then again, I was playing with Dave Treadwell at the time, and he had more masterpoints than most human beings should be entitled to have. Also, our opponents earned their handicaps by merit, so we had no real problem overcoming the deficit.

[manudude03]

I think it should be more multiplicative, but yes, it's approximately true. I think it would be 50/54 x 50% or 46.3%

[WellSpyder]

Fluffy, on 2012-April-24, 08:16, said:

If each field had equal level but then I make a special tournament with mostly good players, is it aproximatelly true that a player with an average 50% put on a field of average 54% would be "expected" to make 46%?

This is exactly the assumption underlying the new EBU National Grading Scheme, discussed recently in this forum.

[Fluffy]

Thx guys.

to Art, I think the same about the final tourney, but right now I am trying to asses if a pair wich is suposed to make 46% or so makes 50%, how good they did and how much does their handicap move because of that. I can asses an 80% handicap to that as well I think and maybe I need to do that in order for the numbers to converge, if I finally gather the data I will tell you.

Another grey point: on a 50% field what is the expected % for a partnership composed of a pro with 57% and a puppil with 47%?

[Fluffy]

manudude03, on 2012-April-24, 08:47, said:

I think it should be more multiplicative, but yes, it's approximately true. I think it would be 50/54 x 50% or 46.3%

Well this took me to realice that a 75% average pair into an average 25% field cannot get an expected 100% I don;t expect the numbers to ever move away from 60% actually, but goes to show that some kind of more complex formula must be taken and adding doesn't work in theory (maybe in practice it does). Your formula is not working either I think (its even worse?)

Maybe some kind of reduction, for example, if you have 60% average your skill gives you 10% extra, if the field is 40% their bad play gives you 10% extra, but you cannot get 2 tops in one board, so when both happen you cannot add more points. This means you cannot just add them together.

[WellSpyder]

Fluffy, on 2012-April-25, 03:30, said:

Another grey point: on a 50% field what is the expected % for a partnership composed of a pro with 57% and a puppil with 47%?

Again, there is a corresponding assumption in the EBU National Grading Scheme, which is that their expected score would be the average of their individual grades, ie 52%.

[mycroft]

According to the ACBL tech files (yes, this is offtopic, but aimed at ArtK78):

The formula for computing the handicap in handicapped knockout team games is:
N x Log10 [(40 + {L/2} + H)/(40 + {L * 3/2})]

where:
N = the number of boards per match
H = the average masterpoint holding of all team members for the team with more masterpoints per member
L = the average masterpoint holding of all team members for the team with fewer masterpoints per member.

[ArtK78]

mycroft, on 2012-April-25, 11:21, said:

According to the ACBL tech files (yes, this is offtopic, but aimed at ArtK78):

The formula for computing the handicap in handicapped knockout team games is:
N x Log10 [(40 + {L/2} + H)/(40 + {L * 3/2})]

where:
N = the number of boards per match
H = the average masterpoint holding of all team members for the team with more masterpoints per member
L = the average masterpoint holding of all team members for the team with fewer masterpoints per member.

Thanks for providing the formula, but you will have to excuse me if I don't commit it to memory.

[mycroft]

One thing that I found interesting = the masterpoint ratio for a particular "handicap per board" is linear; for "1 IMP per board", that ratio is L = 2(H - 360)/29. I'm sure that hrothgar would say "well, that's obvious", but I wasn't thinking about it until I started pulling the graphs.

[Cthulhu D]

http://www.jeff-gold...rg/bridge/study

is an interesting discussion of handicapping too at the end, but only IMP relevant.

[TimG]

Fluffy, on 2012-April-24, 08:16, said:

The question I want to ask is this:

If each field had equal level but then I make a special tournament with mostly good players, is it aproximatelly true that a player with an average 50% put on a field of average 54% would be "expected" to make 46%?

I would expect something different. When a 54% pair faces a 50% pair, I would expect the result to be in a ratio of 54:50. This means that the 54% pair would score about 51.9% and the 50% pair would score about 48.1%.

[Mbodell]

Note there is a similar rating system in US that assumes just that, and which seems to work relatively well to me:

http://www.colorados...PR_FILES/PR.HTM