"Winning chance" usage in the statistical model

[ghost]

Suggestion: what about using the winning chance % from MPGStat to adjust the efficiency?

for example: player.matchs / championshipDays * player.average * (1 + player.goals() * efficiency.coeff * (50%/(1-winning chance) (not sure the math is right).

maybe 1+ln((50%/(1-winning chance)) would be better

Basically, efficiency is unchanged if the winning chance is 50/50. It decreases if eff is lower than 50% and increases if eff is higher than 50%.

Let me know what you think!

Cheers!

[axel3rd]

Thanks for this suggestion !

I'm a little scared this "winning chance" usage falses the model.
Because it will enforce the players of "major teams" in your team.
And in real life, players can have a good notation even if they play in a mid-low team.

But this is just a feeling ... perhaps iterate manually this algorithm on some players could give the truth on this feature.

@mansuydejean : You are a statistics fanatic 😝, WDYT about that ?

[LittleLama]

Hi ! Your fears are justified. Mpgstats winning chances are currently only based on a home madebELO ranking. So I won't suggest to give it an heavy weight in your custom notation. Le jeu. 8 août 2019 à 10:19, Alix Lourme <notifications@github.com> a écrit :

…

Thanks for this suggestion ! I'm a little scared this "winning chance" usage falses the model. Because it will enforce the players of "major teams" in your team. And in real life, players can have a good notation even if they play in a mid-low team. But this is just a feeling ... perhaps iterate manually this algorithm on some players could give the truth on this feature. @mansuydejean <https://github.com/mansuydejean> : You are a statistics fanatic 😝, WDYT about that ? — You are receiving this because you are subscribed to this thread. Reply to this email directly, view it on GitHub <#98?email_source=notifications&email_token=ABGAOD2IYLLJBHD6CCMV2RLQDPJHDA5CNFSM4IKFRIQ2YY3PNVWWK3TUL52HS4DFVREXG43VMVBW63LNMVXHJKTDN5WW2ZLOORPWSZGOD323HLQ#issuecomment-519418798>, or mute the thread <https://github.com/notifications/unsubscribe-auth/ABGAOD3HIHGFUXIW2V2TIVTQDPJHDANCNFSM4IKFRIQQ> .

[ghost]

oh great! thanks for the answer!

[ghost]

Hello!
Thought about it.
Maybe we don't include a winning chance metric to the eff. calculation. We just put it as a new column.
For the winning chance, we parse/scrap it from another source, like a recognized betting website.
What do you think?

[LittleLama]

Two ways to do it : Either each user ask for an API key to the website you chose and integrated to the program. By example : https://www.api-football.com ( https://www.api-football.com/demo/api/v2/odds/league/2 ) Wait for MPGStats to give it : could take 1 or 2 months, maybe less, because not in top of the list. Le lun. 12 août 2019 à 15:19, Nicolas Ballian <notifications@github.com> a écrit :

…

Hello! Thought about it. Maybe we don't include a winning chance metric to the eff. calculation. We just put it as a new column. For the winning chance, we scrap it from another source, like a recognized betting website. What do you think? — You are receiving this because you commented. Reply to this email directly, view it on GitHub <#98?email_source=notifications&email_token=ABGAOD32JP5YGFD6WZZGAZTQEFPI5A5CNFSM4IKFRIQ2YY3PNVWWK3TUL52HS4DFVREXG43VMVBW63LNMVXHJKTDN5WW2ZLOORPWSZGOD4CP4QY#issuecomment-520420931>, or mute the thread <https://github.com/notifications/unsubscribe-auth/ABGAOD4L3I5BXCZFPAONY4DQEFPI5ANCNFSM4IKFRIQQ> .

[axel3rd]

Maybe we don't include a winning chance metric to the eff. calculation. We just put it as a new column.

I promote to stay "simple to use". If a information cannot be included in efficiency formula, it is "useless". Any other columns will reduce the visibility.

Either each user ask for an API key to the website you chose and integrated
to the program.

Always for "keeping simple", adding information from some third-party system with authentication or subscription will complexify the usage.
=> Wait MPGStats information when valuable and includable coherently in efficiency formula is preferalbe.

[LittleLama]

Next match odds are up. Major changes on API, as URL : api.mlnstats.com/leagues/Ligue-1 Please update library so old one can be deleted. Odds are in Ne...o.(home/draw/away) Le lun. 12 août 2019 à 16:29, Alix Lourme <notifications@github.com> a écrit :

…

Maybe we don't include a winning chance metric to the eff. calculation. We just put it as a new column. I promote to stay "simple to use". If a information cannot be included in efficiency formula, it is "useless". Any other columns will reduce the visibility. Either each user ask for an API key to the website you chose and integrated to the program. Always for "keeping simple", adding information from some third-party system with authentication or subscription will complexify the usage. => Wait MPGStats information when valuable and includable coherently in efficiency formula is preferalbe. — You are receiving this because you commented. Reply to this email directly, view it on GitHub <#98?email_source=notifications&email_token=ABGAODYYUWVJA3XGRWEUNZ3QEFXUXA5CNFSM4IKFRIQ2YY3PNVWWK3TUL52HS4DFVREXG43VMVBW63LNMVXHJKTDN5WW2ZLOORPWSZGOD4CWV6A#issuecomment-520448760>, or mute the thread <https://github.com/notifications/unsubscribe-auth/ABGAOD4EBITOAR3QJOBCDGDQEFXUXANCNFSM4IKFRIQQ> .

[axel3rd]

Next match odds are up.
Major changes on API, as URL : api.mlnstats.com/leagues/Ligue-1
Please update library so old one can be deleted.
Odds are in Ne...o.(home/draw/away)

@LittleLama : Many thanks for the information. Will be done in #134.