DEV Community

SoccerDream AI profile picture

SoccerDream AI

404 bio not found

Joined Joined on 
Your table grew 44 columns after row one: auditing schema drift before it rewrites your backtest

Your table grew 44 columns after row one: auditing schema drift before it rewrites your backtest

Comments
7 min read
7,472 write bursts and 48,216 rows: what a prediction log can and cannot prove

7,472 write bursts and 48,216 rows: what a prediction log can and cannot prove

Comments
7 min read
Your leakage audit is reading the wrong clock: 46.6% vs 0.42% on the same 14,781 rows

Your leakage audit is reading the wrong clock: 46.6% vs 0.42% on the same 14,781 rows

Comments
7 min read
96 rows out of 47,114: when 92% of your joins return nothing and 97% of the rest return the wrong match

96 rows out of 47,114: when 92% of your joins return nothing and 97% of the rest return the wrong match

Comments
6 min read
The 8.9% you publish: auditing a visibility gate that is statistically indistinguishable from random

The 8.9% you publish: auditing a visibility gate that is statistically indistinguishable from random

Comments
7 min read
Your 27-model ensemble has 1.5 opinions: measuring diversity with Fleiss' kappa and lineage-weighted voting

Your 27-model ensemble has 1.5 opinions: measuring diversity with Fleiss' kappa and lineage-weighted voting

Comments 1
7 min read
Grade once, prove it forever: 46,756 settled calls and the case for an append-only grading ledger

Grade once, prove it forever: 46,756 settled calls and the case for an append-only grading ledger

Comments 1
6 min read
Your leaderboard is ranking luck: 3,736 models, 46,284 graded calls, and a split-half correlation of 0.04

Your leaderboard is ranking luck: 3,736 models, 46,284 graded calls, and a split-half correlation of 0.04

Comments
7 min read
Deleting your losers buys 4.1 points of hit rate: an audit protocol for model retirement

Deleting your losers buys 4.1 points of hit rate: an audit protocol for model retirement

Comments
7 min read
Data freshness is a schema property: freezing the market price before kickoff

Data freshness is a schema property: freezing the market price before kickoff

Comments
7 min read
The draw is 24% of the outcomes and 0.8% of the labels: auditing class collapse in a three-way model

The draw is 24% of the outcomes and 0.8% of the labels: auditing class collapse in a three-way model

1
Comments 1
7 min read
One number, four answers: reconciling four divergent hit-rate pipelines over 46,139 settled predictions

One number, four answers: reconciling four divergent hit-rate pipelines over 46,139 settled predictions

Comments
7 min read
Your confidence score is a label, not a probability: auditing 45,722 settled predictions

Your confidence score is a label, not a probability: auditing 45,722 settled predictions

Comments
7 min read
Reproducibility is not a seed: what a mutable predictions table really costs you

Reproducibility is not a seed: what a mutable predictions table really costs you

Comments
6 min read
You can't rank 200 models with one leaderboard

You can't rank 200 models with one leaderboard

Comments
6 min read
Immutable predictions: a schema for grading model output you can't quietly rewrite

Immutable predictions: a schema for grading model output you can't quietly rewrite

1
Comments 1
5 min read
loading...