]> git.sesse.net Git - foosball/log
foosball
16 years agoAdjusted initial parameters for maximum prediction power; in particular, master
Steinar H. Gunderson [Wed, 26 Dec 2007 18:05:51 +0000 (19:05 +0100)]
Adjusted initial parameters for maximum prediction power; in particular,
fixed a long-standing off-by-86400 bug.

16 years agoSmall change to make it easier to observe batch runs.
Steinar H. Gunderson [Sun, 9 Dec 2007 21:39:36 +0000 (22:39 +0100)]
Small change to make it easier to observe batch runs.

16 years ago"Fix" a memory leak.
Steinar H. Gunderson [Sun, 9 Dec 2007 15:34:37 +0000 (16:34 +0100)]
"Fix" a memory leak.

16 years agoExtend recalc-single-result.pl to be able to do full batch runs.
Steinar H. Gunderson [Sun, 9 Dec 2007 15:30:18 +0000 (16:30 +0100)]
Extend recalc-single-result.pl to be able to do full batch runs.

16 years agoFix the SWIG startup stuff.
Steinar H. Gunderson [Sun, 9 Dec 2007 15:16:05 +0000 (16:16 +0100)]
Fix the SWIG startup stuff.

16 years agoAdd a SWIG interface, to be able to run recalc-single-result faster.
Steinar H. Gunderson [Sun, 9 Dec 2007 15:14:22 +0000 (16:14 +0100)]
Add a SWIG interface, to be able to run recalc-single-result faster.

16 years agoAbolish non-const references.
Steinar H. Gunderson [Sun, 9 Dec 2007 14:41:27 +0000 (15:41 +0100)]
Abolish non-const references.

16 years agoRemove an obsolete if.
Steinar H. Gunderson [Sun, 9 Dec 2007 14:35:29 +0000 (15:35 +0100)]
Remove an obsolete if.

16 years agoAdd a "dry run" mode to recalc-single-result, for easier testing.
Steinar H. Gunderson [Sun, 9 Dec 2007 14:33:04 +0000 (15:33 +0100)]
Add a "dry run" mode to recalc-single-result, for easier testing.

16 years agoParametrize initial rating/RD instead of hardcoding it.
Steinar H. Gunderson [Sun, 9 Dec 2007 13:33:30 +0000 (14:33 +0100)]
Parametrize initial rating/RD instead of hardcoding it.

16 years agoSwitch to the standard normalization of the logistic pdf. Much more consistent
Steinar H. Gunderson [Sat, 8 Dec 2007 17:24:20 +0000 (18:24 +0100)]
Switch to the standard normalization of the logistic pdf. Much more consistent
with the RDs used by the normal distribution.

16 years agoNow Simpson's rule works like a charm... Odd.
Steinar H. Gunderson [Sat, 8 Dec 2007 16:27:02 +0000 (17:27 +0100)]
Now Simpson's rule works like a charm... Odd.

16 years agoMinor integration tweaks.
Steinar H. Gunderson [Sat, 8 Dec 2007 13:43:00 +0000 (14:43 +0100)]
Minor integration tweaks.

16 years agoImplement proper probabilities (integrating the entire pdf), so assessments
Steinar H. Gunderson [Sat, 8 Dec 2007 12:45:53 +0000 (13:45 +0100)]
Implement proper probabilities (integrating the entire pdf), so assessments
take into account the standard deviation. This also enables more proper model
evaluation, as it mirrors the "predictive discrepancy" used in the Glicko
paper.

16 years agoOptimize the sech²(x) function a bit.
Steinar H. Gunderson [Sun, 2 Dec 2007 14:44:07 +0000 (15:44 +0100)]
Optimize the sech²(x) function a bit.

16 years agoMake the initial rating/RD constants instead of hard-coded values.
Steinar H. Gunderson [Sun, 2 Dec 2007 11:45:59 +0000 (12:45 +0100)]
Make the initial rating/RD constants instead of hard-coded values.

16 years agoSupport the logistic distribution in addition to the normal distribution. Not enabled...
Steinar H. Gunderson [Sun, 2 Dec 2007 11:43:15 +0000 (12:43 +0100)]
Support the logistic distribution in addition to the normal distribution. Not enabled yet.

16 years agoAnother reserve().
Steinar H. Gunderson [Sun, 21 Oct 2007 12:55:07 +0000 (14:55 +0200)]
Another reserve().

16 years agoRemove another mul or so.
Steinar H. Gunderson [Sun, 21 Oct 2007 12:54:11 +0000 (14:54 +0200)]
Remove another mul or so.

16 years agoChange a few divs by constants to muls.
Steinar H. Gunderson [Sun, 21 Oct 2007 12:50:35 +0000 (14:50 +0200)]
Change a few divs by constants to muls.

16 years agoPre-size one vector.
Steinar H. Gunderson [Sun, 21 Oct 2007 12:38:57 +0000 (14:38 +0200)]
Pre-size one vector.

16 years agoGeneralize the Gauss-Jordan solver to arbitrary sizes. In the process it's
Steinar H. Gunderson [Sun, 21 Oct 2007 12:33:34 +0000 (14:33 +0200)]
Generalize the Gauss-Jordan solver to arbitrary sizes. In the process it's
becoming really crufty old code =)

16 years agoMore cleanups, and statify.
Steinar H. Gunderson [Sun, 21 Oct 2007 12:14:40 +0000 (14:14 +0200)]
More cleanups, and statify.

16 years agoUse C++ headers.
Steinar H. Gunderson [Sun, 21 Oct 2007 12:12:09 +0000 (14:12 +0200)]
Use C++ headers.

16 years agoMinor cleanups and optimizations.
Steinar H. Gunderson [Sun, 21 Oct 2007 12:06:08 +0000 (14:06 +0200)]
Minor cleanups and optimizations.

16 years agoMove the games column.
Steinar H. Gunderson [Sun, 21 Oct 2007 11:54:49 +0000 (13:54 +0200)]
Move the games column.

16 years agoFix a few warnings.
Steinar H. Gunderson [Sun, 21 Oct 2007 11:53:28 +0000 (13:53 +0200)]
Fix a few warnings.

16 years agoShow the number of games.
Steinar H. Gunderson [Sun, 21 Oct 2007 11:52:50 +0000 (13:52 +0200)]
Show the number of games.

16 years agoRemoved obsolete (and wrong) function.
Steinar H. Gunderson [Sun, 21 Oct 2007 11:41:45 +0000 (13:41 +0200)]
Removed obsolete (and wrong) function.

16 years agoCentralize rounding decisions, so we round symmetrically.
Steinar H. Gunderson [Sun, 21 Oct 2007 11:40:52 +0000 (13:40 +0200)]
Centralize rounding decisions, so we round symmetrically.

16 years agoConsistency is king.
Steinar H. Gunderson [Sun, 21 Oct 2007 11:34:31 +0000 (13:34 +0200)]
Consistency is king.

16 years agoShorten the headings a bit.
Steinar H. Gunderson [Sun, 21 Oct 2007 11:33:40 +0000 (13:33 +0200)]
Shorten the headings a bit.

16 years ago"Fix" the assessment look too.
Steinar H. Gunderson [Sun, 21 Oct 2007 00:03:21 +0000 (02:03 +0200)]
"Fix" the assessment look too.

16 years agoMade the front page a bit more... different.
Steinar H. Gunderson [Sat, 20 Oct 2007 23:54:34 +0000 (01:54 +0200)]
Made the front page a bit more... different.

16 years agoSupport double assessment, now that it's fast.
Steinar H. Gunderson [Wed, 17 Oct 2007 19:14:02 +0000 (21:14 +0200)]
Support double assessment, now that it's fast.

16 years agoSwitch to FFT-based calculation. Much, much faster!
Steinar H. Gunderson [Wed, 17 Oct 2007 16:39:27 +0000 (18:39 +0200)]
Switch to FFT-based calculation. Much, much faster!

16 years agoFix another bad double bug...
Steinar H. Gunderson [Wed, 17 Oct 2007 16:36:41 +0000 (18:36 +0200)]
Fix another bad double bug...

16 years agoFix the ordering of almost-equal games on the front page.
Steinar H. Gunderson [Tue, 16 Oct 2007 15:22:08 +0000 (17:22 +0200)]
Fix the ordering of almost-equal games on the front page.

16 years agoTypofix.
Steinar H. Gunderson [Thu, 11 Oct 2007 19:30:00 +0000 (21:30 +0200)]
Typofix.

16 years agoInvert the sign in the inner integral, as it's then expressible as a convolution.
Steinar H. Gunderson [Thu, 11 Oct 2007 19:28:34 +0000 (21:28 +0200)]
Invert the sign in the inner integral, as it's then expressible as a convolution.

16 years agoSwitch to more accurate double score calculation.
Steinar H. Gunderson [Thu, 11 Oct 2007 17:06:25 +0000 (19:06 +0200)]
Switch to more accurate double score calculation.

16 years agoPrecompute the binomial as a single value instead of two separate ones.
Steinar H. Gunderson [Wed, 10 Oct 2007 22:04:28 +0000 (00:04 +0200)]
Precompute the binomial as a single value instead of two separate ones.

16 years agoMake a common function for the Simpson integration stuff.
Steinar H. Gunderson [Wed, 10 Oct 2007 22:01:24 +0000 (00:01 +0200)]
Make a common function for the Simpson integration stuff.

16 years agoFix syntax error on new double results.
Steinar H. Gunderson [Wed, 10 Oct 2007 21:46:28 +0000 (23:46 +0200)]
Fix syntax error on new double results.

16 years agoRemove a double negative.
Steinar H. Gunderson [Mon, 8 Oct 2007 18:47:50 +0000 (20:47 +0200)]
Remove a double negative.

16 years agoSupport k != 10, ie. the new table.
Steinar H. Gunderson [Mon, 8 Oct 2007 18:15:05 +0000 (20:15 +0200)]
Support k != 10, ie. the new table.

16 years agoFixed insertion of new results again.
Steinar H. Gunderson [Mon, 8 Oct 2007 11:15:31 +0000 (13:15 +0200)]
Fixed insertion of new results again.

16 years agoMore writing about FoosRank.
Steinar H. Gunderson [Fri, 5 Oct 2007 21:34:08 +0000 (23:34 +0200)]
More writing about FoosRank.

16 years agoChanged a few headings.
Steinar H. Gunderson [Fri, 5 Oct 2007 21:31:31 +0000 (23:31 +0200)]
Changed a few headings.

16 years agoShow the rating difference for each game.
Steinar H. Gunderson [Fri, 5 Oct 2007 21:27:57 +0000 (23:27 +0200)]
Show the rating difference for each game.

16 years agoShow the ratings in the assessments.
Steinar H. Gunderson [Fri, 5 Oct 2007 21:05:00 +0000 (23:05 +0200)]
Show the ratings in the assessments.

16 years agofoorank -> foosrank
Steinar H. Gunderson [Fri, 5 Oct 2007 20:57:53 +0000 (22:57 +0200)]
foorank -> foosrank

16 years agoBug fixes.
Steinar H. Gunderson [Fri, 5 Oct 2007 16:45:52 +0000 (18:45 +0200)]
Bug fixes.

16 years agoOops, also remember the C++ part of the assessment.
Steinar H. Gunderson [Fri, 5 Oct 2007 16:45:39 +0000 (18:45 +0200)]
Oops, also remember the C++ part of the assessment.

16 years agoAdded an assessment function.
Steinar H. Gunderson [Fri, 5 Oct 2007 16:45:11 +0000 (18:45 +0200)]
Added an assessment function.

16 years agoSplit out the team rating calculation into a separate function.
Steinar H. Gunderson [Fri, 5 Oct 2007 16:43:56 +0000 (18:43 +0200)]
Split out the team rating calculation into a separate function.

16 years agoSplit out a function.
Steinar H. Gunderson [Wed, 3 Oct 2007 23:17:02 +0000 (01:17 +0200)]
Split out a function.

16 years agoSwitch to using Simpson's rule for the moment estimation as well, which seems
Steinar H. Gunderson [Wed, 3 Oct 2007 23:11:06 +0000 (01:11 +0200)]
Switch to using Simpson's rule for the moment estimation as well, which seems
to yield a way better initial estimate.

16 years agoSwitch to using Simpson's rule for the integration, and discouple the two
Steinar H. Gunderson [Wed, 3 Oct 2007 22:07:11 +0000 (00:07 +0200)]
Switch to using Simpson's rule for the integration, and discouple the two
step sizes.

16 years agoRemove obsolete code.
Steinar H. Gunderson [Wed, 3 Oct 2007 21:33:25 +0000 (23:33 +0200)]
Remove obsolete code.

16 years agoMake the recalc stuff age the RDs properly.
Steinar H. Gunderson [Wed, 3 Oct 2007 21:31:41 +0000 (23:31 +0200)]
Make the recalc stuff age the RDs properly.

16 years agoMake a constant out of 455.
Steinar H. Gunderson [Wed, 3 Oct 2007 21:23:53 +0000 (23:23 +0200)]
Make a constant out of 455.

16 years agoFoosRank commercials.
Steinar H. Gunderson [Wed, 3 Oct 2007 21:20:13 +0000 (23:20 +0200)]
FoosRank commercials.

16 years agoDon't show ratings without any information.
Steinar H. Gunderson [Wed, 3 Oct 2007 21:16:18 +0000 (23:16 +0200)]
Don't show ratings without any information.

16 years agoFix the initial sigma estimate.
Steinar H. Gunderson [Wed, 3 Oct 2007 21:15:13 +0000 (23:15 +0200)]
Fix the initial sigma estimate.

16 years agoSwitch to FooRank!
Steinar H. Gunderson [Wed, 3 Oct 2007 20:44:46 +0000 (22:44 +0200)]
Switch to FooRank!

16 years agoAdd recalc scripts.
Steinar H. Gunderson [Wed, 3 Oct 2007 20:35:20 +0000 (22:35 +0200)]
Add recalc scripts.

16 years agoFixed a bad bug...
Steinar H. Gunderson [Wed, 3 Oct 2007 20:19:33 +0000 (22:19 +0200)]
Fixed a bad bug...

16 years agoRemove superfluous stuff.
Steinar H. Gunderson [Wed, 3 Oct 2007 20:03:55 +0000 (22:03 +0200)]
Remove superfluous stuff.

16 years agoInitial checkin.
Steinar H. Gunderson [Wed, 3 Oct 2007 19:59:01 +0000 (21:59 +0200)]
Initial checkin.