Novel Market Foresight Premium

mlb: MFP rises from 0.06 to 0.24 over the game (mean 0.25); soccer_intl: MFP rises from 0.64 to 1.71 over the game (mean 0.92)

View full size ↗

confirmednull (a finding)not testabledescriptivepending

Chart: Novel Market Foresight Premium -- mlb: MFP rises from 0.06 to 0.24 over the game (mean 0.25); soccer_intl: MFP rises from 0.64 to 1.71 over the game (mean 0.92) — scripts/platformkit/analytics_showcase/out/novel_market_foresight_premium.json

Market Foresight PremiumMFP

mlb: MFP rises from 0.06 to 0.24 over the game (mean 0.25); soccer_intl: MFP rises from 0.64 to 1.71 over the game (mean 0.92)

What it measures

The market's excess resolving power over a scoreboard-only model, per bit of uncertainty still left in the game, by checkpoint.

The formula

MFP(t) = ((naive-market)/naive - (naive-model)/naive) / max(entropy_market_bits, 0.15)

The results

mlb

n checkpoints 10mfp first 0.063mfp last 0.2365mfp mean 0.2461mfp delta last minus first 0.1735

checkpoint	n	market skill	model skill	entropy market bits	entropy floored	mfp
1	7646	0.013	-0.0479	0.9669	no	0.063
2	5921	0.0619	-0.0272	0.9259	no	0.0963
3	6333	0.0619	-0.0205	0.8731	no	0.0944
4	6266	0.066	-0.0289	0.8417	no	0.1127
5	6021	0.1074	-0.0184	0.7923	no	0.1588
6	6478	0.1664	-0.0253	0.7629	no	0.2513
7	6311	0.233	-0.0689	0.7219	no	0.4182
8	5276	0.3141	-0.0409	0.6979	no	0.5086
9	2282	0.3254	-0.0519	0.7244	no	0.5208
10	94	0.0146	-0.1926	0.8761	no	0.2365

soccer_intl

n checkpoints 19mfp first 0.6393mfp last 1.7056mfp mean 0.9211mfp delta last minus first 1.0663

checkpoint	n	market skill	model skill	entropy market bits	entropy floored	mfp
0	165	0.4649	-0.0599	0.8209	no	0.6393
5	238	0.4422	-0.1471	0.8399	no	0.7016
10	224	0.4624	-0.1698	0.8363	no	0.7559
15	233	0.4191	-0.2145	0.8082	no	0.784
20	215	0.3729	-0.1946	0.8275	no	0.6858
25	218	0.2859	-0.2676	0.8131	no	0.6807
30	219	0.2424	-0.3711	0.7788	no	0.7877
35	198	0.2194	-0.3224	0.7556	no	0.7171
40	211	0.2018	-0.4252	0.7406	no	0.8466
45	429	0.313	-0.2648	0.744	no	0.7766
50	197	-0.0278	-0.6128	0.6522	no	0.897
55	202	-0.0718	-0.5369	0.616	no	0.7549
60	194	-0.0404	-0.5287	0.6276	no	0.7781
65	156	-0.1567	-0.7093	0.6637	no	0.8325
70	154	-0.352	-0.9791	0.6739	no	0.9304
75	136	-0.1634	-0.9658	0.7377	no	1.0877
80	131	-0.2661	-1.1949	0.6955	no	1.3354
85	101	-0.522	-1.5904	0.5922	no	1.8041
90	37	-0.3982	-1.1476	0.4394	no	1.7056

Prior art

INCREMENTAL

We searched for prior work before claiming anything. Here is what exists:

The absorption instrument (paired model/market/naive Brier by game-time) is arXiv 2606.07811 ('When Do Markets Fully Process Public Information', Kalshi/NBA), which info_arrival_curve.json's own novelty block flags INCREMENTAL. Entropy decay is textbook (Shannon). Do NOT claim first-ever on the components -- only the skill-gap-PER-BIT normalization (the unpublished cross of the two separate literatures) is claimed.

Declared confounds

'model' is one specific state-only forecaster, so MFP conflates genuine news the market prices (lineups, momentum) with plain model misspecification.
Entropy -> 0 late in a game would explode the ratio; the 0.15-bit floor caps that.
mlb (inning) and soccer_intl (5-min bucket) only; different clock units, not equated.

scripts/platformkit/analytics_showcase/out/info_arrival_curve.json

scripts/platformkit/analytics_showcase/out/market_convergence.json

edge_claimed: false · descriptive_only

Ask Scout about this

Where does the in-game market know the most that the scoreboard doesn't?Compare how MLB and soccer markets converge during a game.Compare model performance at the very start versus the very end of a game.Ask anything →