Sizing-Criteria Analysis · V5 baseline

Full 10-asset universe · IS 600 trades · OOS 315 trades · conservative costs · generated 2026-06-12 23:13 UTC

What this is

V5 entries/exits are unchanged. Each trade is tagged with state known at entry — only asset-agnostic features that travel with the daily-rotating live universe (no asset identity). We then ask: which conditions predict a better R-multiple? That is the empirical size-up / size-down map.

Outcome = R-multiple (PnL ÷ risk-at-entry), size-independent. "Implied size mult" in the JSON is descriptive only — any rule must be backtested as a single overlay before shipping. Effects are validated by sign-persistence IS→OOS and a Welch t-test below.

Headline contrasts — significance (Welch t-test on R-multiples)

|t| ≥ 1.96 ≈ p<0.05. Directional persistence alone is not enough — this is the honest bar.

period	contrast	mean-R (A)	mean-R (B)	t	n A/B
IS	Volatility: mid-band vs rest	0.430	0.536	-0.45	106/494	not sig.
IS	Portfolio heat: 2 open vs ≤1 open	0.606	0.467	0.80	217/383	not sig.
IS	Group: correlated-open vs first-in-group	0.688	0.483	0.95	101/499	not sig.
IS	Side: SHORT vs LONG	0.403	0.574	-1.06	200/400	not sig.
OOS	Volatility: mid-band vs rest	-0.104	0.542	-2.49	63/252	SIGNIFICANT
OOS	Portfolio heat: 2 open vs ≤1 open	0.620	0.292	1.46	116/199	not sig.
OOS	Group: correlated-open vs first-in-group	0.619	0.382	0.76	41/274	not sig.
OOS	Side: SHORT vs LONG	0.361	0.425	-0.30	60/255	not sig.

Portfolio heat (concurrent open positions at entry)

population mean R — IS 0.517 · OOS 0.413

bucket	IS n	IS win	IS mean-R	OOS n	OOS win	OOS mean-R
heat:0concurrent	180	67.2%	0.623	97	58.8%	0.225
heat:2concurrent	217	62.2%	0.606	116	65.5%	0.620
heat:1concurrent	203	59.1%	0.329	102	58.8%	0.355

Correlation group occupancy at entry

population mean R — IS 0.517 · OOS 0.413

bucket	IS n	IS win	IS mean-R	OOS n	OOS win	OOS mean-R
grp:correlated_open	101	60.4%	0.688	41	70.7%	0.619
grp:first_in_group	499	63.1%	0.483	274	59.9%	0.382

Volatility state (asset-relative ATR percentile)

population mean R — IS 0.517 · OOS 0.413

bucket	IS n	IS win	IS mean-R	OOS n	OOS win	OOS mean-R
vol:low	45	66.7%	0.747	22	72.7%	0.474
vol:high	449	62.1%	0.515	230	64.3%	0.548
vol:mid	106	63.2%	0.430	63	46.0%	-0.104

Setup quality (RR headroom over threshold)

population mean R — IS 0.517 · OOS 0.413

bucket	IS n	IS win	IS mean-R	OOS n	OOS win	OOS mean-R
setup:weak	269	66.5%	0.597	123	61.0%	0.338
setup:medium	231	61.9%	0.548	147	59.9%	0.424
setup:strong	100	54.0%	0.232	45	66.7%	0.582

Macro regime

population mean R — IS 0.517 · OOS 0.413

bucket	IS n	IS win	IS mean-R	OOS n	OOS win	OOS mean-R
regime:ACCUMULATION	58	60.3%	1.119	—	—	—
regime:BEAR	178	62.4%	0.461	60	63.3%	0.361
regime:BULL	364	63.2%	0.449	255	60.8%	0.425

Session (UTC hour band)

population mean R — IS 0.517 · OOS 0.413

bucket	IS n	IS win	IS mean-R	OOS n	OOS win	OOS mean-R
sess:eu_morning	244	61.9%	0.549	90	62.2%	0.427
sess:asia_offhours	93	68.8%	0.537	58	63.8%	0.441
sess:us	263	61.2%	0.481	167	59.9%	0.395

Trade side

population mean R — IS 0.517 · OOS 0.413

bucket	IS n	IS win	IS mean-R	OOS n	OOS win	OOS mean-R
side:LONG	400	64.5%	0.575	255	60.8%	0.425
side:SHORT	200	59.0%	0.404	60	63.3%	0.361

Entry kind

population mean R — IS 0.517 · OOS 0.413

bucket	IS n	IS win	IS mean-R	OOS n	OOS win	OOS mean-R
kind:PULLBACK	600	62.7%	0.517	315	61.3%	0.413