- Calculating odds from trueskill ratings? by Nick Bonfatti
- Re: [trueskill] Calculating odds from trueskill ratings? by Monty Paython
- Re: [trueskill] Calculating odds from trueskill ratings? by Heungsub Lee
- Re: [trueskill] Calculating odds from trueskill ratings? by Nick Bonfatti
- Re: [trueskill] Calculating odds from trueskill ratings? by Nick Bonfatti
- Re: [trueskill] Calculating odds from trueskill ratings? by Heungsub Lee
- Re: [trueskill] Calculating odds from trueskill ratings? by Nick Bonfatti
- Re: [trueskill] Calculating odds from trueskill ratings? by Heungsub Lee
- Re: [trueskill] Calculating odds from trueskill ratings? by Nick Bonfatti

- From:
- Nick Bonfatti
- Date:
- 2013-08-19 @ 19:27

I'm interested in (just for fun) calculating odds from comparing two trueskill ratings. I know I can get the draw probability, and I'm guessing there's a related method to determine the odds of one side winning. (1- draw probability) gives us the probability someone will definitively win, then we need to split that into the odds for one side vs the other. But, how to determine the split? I'm not great at statistics so I wouldn't be surprised if I was completely off-base.

- From:
- Monty Paython
- Date:
- 2013-08-20 @ 01:17

Hi Nick, Maybe our first issue (https://github.com/sublee/trueskill/issues/1) could help you. Is it you are finding? Regards, Heungsub 2013/8/20 Nick Bonfatti <nick@bonfatti.net> > I'm interested in (just for fun) calculating odds from comparing two > trueskill ratings. I know I can get the draw probability, and I'm guessing > there's a related method to determine the odds of one side winning. (1- > draw probability) gives us the probability someone will definitively win, > then we need to split that into the odds for one side vs the other. But, > how to determine the split? I'm not great at statistics so I wouldn't be > surprised if I was completely off-base. >

- From:
- Heungsub Lee
- Date:
- 2013-08-20 @ 01:21

Sorry, my real email address is this not "Monty Paython". There was a confusion. 2013/8/20 Monty Paython <monty@paython.org> > Hi Nick, > > Maybe our first issue (https://github.com/sublee/trueskill/issues/1) > could help you. Is it you are finding? > > Regards, > Heungsub > > > 2013/8/20 Nick Bonfatti <nick@bonfatti.net> > >> I'm interested in (just for fun) calculating odds from comparing two >> trueskill ratings. I know I can get the draw probability, and I'm guessing >> there's a related method to determine the odds of one side winning. (1- >> draw probability) gives us the probability someone will definitively win, >> then we need to split that into the odds for one side vs the other. But, >> how to determine the split? I'm not great at statistics so I wouldn't be >> surprised if I was completely off-base. >> > >

- From:
- Nick Bonfatti
- Date:
- 2013-08-20 @ 03:18

Thanks, I'll check that out. On Mon, Aug 19, 2013 at 9:21 PM, Heungsub Lee <sub@subl.ee> wrote: > Sorry, my real email address is this not "Monty Paython". There was a > confusion. > > > 2013/8/20 Monty Paython <monty@paython.org> > >> Hi Nick, >> >> Maybe our first issue (https://github.com/sublee/trueskill/issues/1) >> could help you. Is it you are finding? >> >> Regards, >> Heungsub >> >> >> 2013/8/20 Nick Bonfatti <nick@bonfatti.net> >> >>> I'm interested in (just for fun) calculating odds from comparing two >>> trueskill ratings. I know I can get the draw probability, and I'm guessing >>> there's a related method to determine the odds of one side winning. (1- >>> draw probability) gives us the probability someone will definitively win, >>> then we need to split that into the odds for one side vs the other. But, >>> how to determine the split? I'm not great at statistics so I wouldn't be >>> surprised if I was completely off-base. >>> >> >> >

- From:
- Nick Bonfatti
- Date:
- 2013-08-20 @ 03:38

OK so that gives (a reasonable approximation and good starting point anyway) the probability of a win for a 1 vs 1 match, but how bout n vs n? My project involves teams of 3. Also, unrelated: I save the rating values to a db. Should I save the environment as well, or is it OK to make a new environment every time I run my script? On Mon, Aug 19, 2013 at 11:18 PM, Nick Bonfatti <nick@bonfatti.net> wrote: > Thanks, I'll check that out. > > > On Mon, Aug 19, 2013 at 9:21 PM, Heungsub Lee <sub@subl.ee> wrote: > >> Sorry, my real email address is this not "Monty Paython". There was a >> confusion. >> >> >> 2013/8/20 Monty Paython <monty@paython.org> >> >>> Hi Nick, >>> >>> Maybe our first issue (https://github.com/sublee/trueskill/issues/1) >>> could help you. Is it you are finding? >>> >>> Regards, >>> Heungsub >>> >>> >>> 2013/8/20 Nick Bonfatti <nick@bonfatti.net> >>> >>>> I'm interested in (just for fun) calculating odds from comparing two >>>> trueskill ratings. I know I can get the draw probability, and I'm guessing >>>> there's a related method to determine the odds of one side winning. (1- >>>> draw probability) gives us the probability someone will definitively win, >>>> then we need to split that into the odds for one side vs the other. But, >>>> how to determine the split? I'm not great at statistics so I wouldn't be >>>> surprised if I was completely off-base. >>>> >>> >>> >> >

- From:
- Heungsub Lee
- Date:
- 2013-08-21 @ 00:05

Do you save a pickled <class trueskill.TrueSkill> object to the DB? 2013/8/20 Nick Bonfatti <nick@bonfatti.net> > OK so that gives (a reasonable approximation and good starting point > anyway) the probability of a win for a 1 vs 1 match, but how bout n vs n? > My project involves teams of 3. > > Also, unrelated: I save the rating values to a db. Should I save the > environment as well, or is it OK to make a new environment every time I run > my script? > > > On Mon, Aug 19, 2013 at 11:18 PM, Nick Bonfatti <nick@bonfatti.net> wrote: > >> Thanks, I'll check that out. >> >> >> On Mon, Aug 19, 2013 at 9:21 PM, Heungsub Lee <sub@subl.ee> wrote: >> >>> Sorry, my real email address is this not "Monty Paython". There was a >>> confusion. >>> >>> >>> 2013/8/20 Monty Paython <monty@paython.org> >>> >>>> Hi Nick, >>>> >>>> Maybe our first issue (https://github.com/sublee/trueskill/issues/1) >>>> could help you. Is it you are finding? >>>> >>>> Regards, >>>> Heungsub >>>> >>>> >>>> 2013/8/20 Nick Bonfatti <nick@bonfatti.net> >>>> >>>>> I'm interested in (just for fun) calculating odds from comparing two >>>>> trueskill ratings. I know I can get the draw probability, and I'm guessing >>>>> there's a related method to determine the odds of one side winning. (1- >>>>> draw probability) gives us the probability someone will definitively win, >>>>> then we need to split that into the odds for one side vs the other. But, >>>>> how to determine the split? I'm not great at statistics so I wouldn't be >>>>> surprised if I was completely off-base. >>>>> >>>> >>>> >>> >> >

- From:
- Nick Bonfatti
- Date:
- 2013-08-21 @ 00:37

I've just been instantiating a new one every time, it only just occurred to me that it might be a good idea to use a consistent environment too. On Tue, Aug 20, 2013 at 8:05 PM, Heungsub Lee <sub@subl.ee> wrote: > Do you save a pickled <class trueskill.TrueSkill> object to the DB? > > > 2013/8/20 Nick Bonfatti <nick@bonfatti.net> > >> OK so that gives (a reasonable approximation and good starting point >> anyway) the probability of a win for a 1 vs 1 match, but how bout n vs n? >> My project involves teams of 3. >> >> Also, unrelated: I save the rating values to a db. Should I save the >> environment as well, or is it OK to make a new environment every time I run >> my script? >> >> >> On Mon, Aug 19, 2013 at 11:18 PM, Nick Bonfatti <nick@bonfatti.net>wrote: >> >>> Thanks, I'll check that out. >>> >>> >>> On Mon, Aug 19, 2013 at 9:21 PM, Heungsub Lee <sub@subl.ee> wrote: >>> >>>> Sorry, my real email address is this not "Monty Paython". There was a >>>> confusion. >>>> >>>> >>>> 2013/8/20 Monty Paython <monty@paython.org> >>>> >>>>> Hi Nick, >>>>> >>>>> Maybe our first issue (https://github.com/sublee/trueskill/issues/1) >>>>> could help you. Is it you are finding? >>>>> >>>>> Regards, >>>>> Heungsub >>>>> >>>>> >>>>> 2013/8/20 Nick Bonfatti <nick@bonfatti.net> >>>>> >>>>>> I'm interested in (just for fun) calculating odds from comparing two >>>>>> trueskill ratings. I know I can get the draw probability, and I'm guessing >>>>>> there's a related method to determine the odds of one side winning. (1- >>>>>> draw probability) gives us the probability someone will definitively win, >>>>>> then we need to split that into the odds for one side vs the other. But, >>>>>> how to determine the split? I'm not great at statistics so I wouldn't be >>>>>> surprised if I was completely off-base. >>>>>> >>>>> >>>>> >>>> >>> >> >

- From:
- Heungsub Lee
- Date:
- 2013-08-21 @ 00:43

Do like this: env = trueskill.TrueSkill(MY_MU, MY_SIGMA) You don't need to do like this: env = pickle.loads(db_load('trueskill')) 2013/8/21 Nick Bonfatti <nick@bonfatti.net> > I've just been instantiating a new one every time, it only just occurred > to me that it might be a good idea to use a consistent environment too. > > > On Tue, Aug 20, 2013 at 8:05 PM, Heungsub Lee <sub@subl.ee> wrote: > >> Do you save a pickled <class trueskill.TrueSkill> object to the DB? >> >> >> 2013/8/20 Nick Bonfatti <nick@bonfatti.net> >> >>> OK so that gives (a reasonable approximation and good starting point >>> anyway) the probability of a win for a 1 vs 1 match, but how bout n vs n? >>> My project involves teams of 3. >>> >>> Also, unrelated: I save the rating values to a db. Should I save the >>> environment as well, or is it OK to make a new environment every time I run >>> my script? >>> >>> >>> On Mon, Aug 19, 2013 at 11:18 PM, Nick Bonfatti <nick@bonfatti.net>wrote: >>> >>>> Thanks, I'll check that out. >>>> >>>> >>>> On Mon, Aug 19, 2013 at 9:21 PM, Heungsub Lee <sub@subl.ee> wrote: >>>> >>>>> Sorry, my real email address is this not "Monty Paython". There was a >>>>> confusion. >>>>> >>>>> >>>>> 2013/8/20 Monty Paython <monty@paython.org> >>>>> >>>>>> Hi Nick, >>>>>> >>>>>> Maybe our first issue (https://github.com/sublee/trueskill/issues/1) >>>>>> could help you. Is it you are finding? >>>>>> >>>>>> Regards, >>>>>> Heungsub >>>>>> >>>>>> >>>>>> 2013/8/20 Nick Bonfatti <nick@bonfatti.net> >>>>>> >>>>>>> I'm interested in (just for fun) calculating odds from comparing two >>>>>>> trueskill ratings. I know I can get the draw probability, and I'm guessing >>>>>>> there's a related method to determine the odds of one side winning. (1- >>>>>>> draw probability) gives us the probability someone will definitively win, >>>>>>> then we need to split that into the odds for one side vs the other. But, >>>>>>> how to determine the split? I'm not great at statistics so I wouldn't be >>>>>>> surprised if I was completely off-base. >>>>>>> >>>>>> >>>>>> >>>>> >>>> >>> >> >

- From:
- Nick Bonfatti
- Date:
- 2013-08-21 @ 01:34

OK that's what I've been doing. Thanks! On Tue, Aug 20, 2013 at 8:43 PM, Heungsub Lee <sub@subl.ee> wrote: > Do like this: > > env = trueskill.TrueSkill(MY_MU, MY_SIGMA) > > You don't need to do like this: > > env = pickle.loads(db_load('trueskill')) > > > 2013/8/21 Nick Bonfatti <nick@bonfatti.net> > >> I've just been instantiating a new one every time, it only just occurred >> to me that it might be a good idea to use a consistent environment too. >> >> >> On Tue, Aug 20, 2013 at 8:05 PM, Heungsub Lee <sub@subl.ee> wrote: >> >>> Do you save a pickled <class trueskill.TrueSkill> object to the DB? >>> >>> >>> 2013/8/20 Nick Bonfatti <nick@bonfatti.net> >>> >>>> OK so that gives (a reasonable approximation and good starting point >>>> anyway) the probability of a win for a 1 vs 1 match, but how bout n vs n? >>>> My project involves teams of 3. >>>> >>>> Also, unrelated: I save the rating values to a db. Should I save the >>>> environment as well, or is it OK to make a new environment every time I run >>>> my script? >>>> >>>> >>>> On Mon, Aug 19, 2013 at 11:18 PM, Nick Bonfatti <nick@bonfatti.net>wrote: >>>> >>>>> Thanks, I'll check that out. >>>>> >>>>> >>>>> On Mon, Aug 19, 2013 at 9:21 PM, Heungsub Lee <sub@subl.ee> wrote: >>>>> >>>>>> Sorry, my real email address is this not "Monty Paython". There was a >>>>>> confusion. >>>>>> >>>>>> >>>>>> 2013/8/20 Monty Paython <monty@paython.org> >>>>>> >>>>>>> Hi Nick, >>>>>>> >>>>>>> Maybe our first issue (https://github.com/sublee/trueskill/issues/1) >>>>>>> could help you. Is it you are finding? >>>>>>> >>>>>>> Regards, >>>>>>> Heungsub >>>>>>> >>>>>>> >>>>>>> 2013/8/20 Nick Bonfatti <nick@bonfatti.net> >>>>>>> >>>>>>>> I'm interested in (just for fun) calculating odds from comparing >>>>>>>> two trueskill ratings. I know I can get the draw probability, and I'm >>>>>>>> guessing there's a related method to determine the odds of one side >>>>>>>> winning. (1- draw probability) gives us the probability someone will >>>>>>>> definitively win, then we need to split that into the odds for one side vs >>>>>>>> the other. But, how to determine the split? I'm not great at statistics so >>>>>>>> I wouldn't be surprised if I was completely off-base. >>>>>>>> >>>>>>> >>>>>>> >>>>>> >>>>> >>>> >>> >> >