22, Multi-market backtest results
Generated: 2026-05-16T21:33:25.515Z · Wall clock: 134s
Per-market BUY-only return statistics
| Market |
Total decls |
With BT row |
Priced |
T+30 mean / WR / n |
T+90 mean / WR / n / t-stat / Sharpe |
T+180 mean / WR / n |
T+365 mean / WR / n |
| AMF |
25865 |
24116 |
23076 |
0.20% / 49% / 16534 |
-0.32% / 44% / 16005 / t=-1.38 / Sh=-0.011 |
-0.11% / 43% / 15391 |
4.47% / 44% / 13486 |
| SEC |
6206 |
5718 |
5696 |
7.12% / 63% / 237 |
6.16% / 60% / 202 / t=2.85 / Sh=0.201 |
10.69% / 60% / 164 |
-1.17% / 49% / 69 |
| BAFIN |
295 |
40 |
40 |
0.28% / 46% / 24 |
1.30% / 38% / 24 / t=0.25 / Sh=0.052 |
-0.27% / 42% / 24 |
, |
| SIX |
226 |
2 |
2 |
, |
, |
, |
, |
| RNS |
176 |
172 |
166 |
, |
, |
, |
, |
| SEDI |
68 |
41 |
38 |
, |
, |
, |
, |
Combined universe (equal-weight all signals)
- R30: mean 0.30% · WR 49% · n=16795 · t=2.59 · Sharpe=0.020
- R90: mean -0.24% · WR 44% · n=16231 · t=-1.03 · Sharpe=-0.008
- R160: mean 0.00% · WR 43% · n=15579 · t=0.00 · Sharpe=0.000
- R365: mean 4.44% · WR 44% · n=13555 · t=2.44 · Sharpe=0.021
Per-market vs AMF delta @ T+90
| Market |
T+90 mean |
Δ vs AMF (-0.32%) |
WR |
n |
| AMF |
-0.32% |
+0.00pp |
44% |
16005 |
| SEC |
6.16% |
+6.48pp |
60% |
202 |
| BAFIN |
1.30% |
+1.63pp |
38% |
24 |
| SIX |
, |
, |
, |
0 |
| RNS |
, |
, |
, |
0 |
| SEDI |
, |
, |
, |
0 |
Markets ranked by Sharpe @ T+90 (n≥3)
- SEC, Sharpe 0.201 · mean 6.16% · WR 60% · n=202
- BAFIN, Sharpe 0.052 · mean 1.30% · WR 38% · n=24
- AMF, Sharpe -0.011 · mean -0.32% · WR 44% · n=16005
Symbol-coverage per market
| Market |
Companies (total) |
Companies resolved |
Coverage % |
n decls priced |
| AMF |
601 |
514 |
86% |
23076 |
| SEC |
1687 |
1687 |
100% |
5696 |
| BAFIN |
291 |
38 |
13% |
40 |
| SIX |
65 |
1 |
2% |
2 |
| RNS |
56 |
56 |
100% |
166 |
| SEDI |
21 |
21 |
100% |
38 |
Scoring weight tuning proposals (hypotheses, not applied)
Based on per-market Sharpe @ T+90 vs AMF baseline:
- SEC (n=202, mean 6.16%, WR 60%): apply +0.06× score boost on SEC filings (Sharpe stronger by 0.21 pts)
- BAFIN (n=24, mean 1.30%, WR 38%): apply +0.02× score boost on BAFIN filings (Sharpe stronger by 0.06 pts)
Caveats per market
- AMF: full coverage (2020-03 to 2026-05). Reference baseline.
- SEC: 6206 decls / 1y backfill window. Direction classifier handles French nature labels stored on SEC rows (legacy).
Company.market already contains the ticker → 1:1 Yahoo symbol; no ISIN needed.
- BAFIN: tiny n (295), only top 30-50 issuers resolved (ISIN→Yahoo search). Remaining excluded as 'no-symbol'.
- SIX: 226 decls, anonymised insider names → no track-record component possible; small n; same top-50 ISIN cap.
- RNS: 176 decls, ticker→
.L cheap mapping. Window very short (4 days at time of run).
- SEDI: 68 decls, ticker→
.TO mapping; numeric SEDI nature codes (e.g. "10 - Acquisition") need explicit recognition added to direction classifier.
- All non-AMF markets: backfill windows are recent (≤1y for SEC, ≤14d for BAFIN/SIX/RNS/SEDI). T+365/T+180 metrics for non-AMF are effectively null or based on a handful of older outliers. Use T+30/T+90 as primary comparators.