DC Market Intelligence Report –Sep 13, 2026

01 The Market Verdict

Yesterday’s strong-steamer cohort underperformed the win expectation at final prices. The group produced 11 winners from 90 qualifiers (12.2% win rate) against 12.49 expected, giving a performance index of 88.1. That is not a collapse in signal, but it is clear evidence that—on the day—late contraction was not reliably translating into wins at the rate the closing market implied.

The more important context is that this is not isolated noise. Over the last 30 days (2,256 qualifiers), the profile is similar: 317 winners versus 362.38 expected, performance index 87.5. In other words, the closing-price “probability” attached to these strong steamers has been running hot relative to realised outcomes for a sustained period. That does not mean every steamer is wrong; it means the subset defined here (≥15% contraction from 10am to final) has, on aggregate, been overbet at the close.

The P/L split reinforces what that overbetting looks like in practice. Theoretical level-stakes P/L at 10am is negative yesterday (-12.74 pts) but massively positive over 30 days (+496.93 pts), while the SP line is negative both yesterday (-35.54 pts) and over 30 days (-373.05 pts). The market evidence is that the edge—where it exists—has been earlier in the day, with value largely given back (and more) by the time the crowd has finished compressing prices late.

02 Market Beaters vs Money That Misfired

03 Where the Money Went

04 The Practical Read

1. Use late steam as a confirmation layer, not a decision engine. With a one-day PI of 88.1 and a 30-day PI of 87.5, the closing market has been pricing this subset as more likely to win than results have supported.
2. Separate “looks dramatic” from “adds probability.” Huge percentage contractions at long odds (e.g., 50/1 into teens) can be eye-catching while still leaving the horse with a low implied chance; treat these as information, but keep expectations proportional.
3. Respect places as signal quality, but price them correctly. Yesterday’s 34.4% place rate shows the market often identified contenders; the issue is that final prices implied more outright winning than materialised.
4. Track where the edge sits in time. The 30-day split—positive at 10am (+496.93 pts) and negative at SP (-373.05 pts)—is consistent with value being earlier and being eroded by late compression. Watch whether the late move is merely “catch-up” to a price that was already value earlier.
5. Interpret pp movement in context. The average Open→10am move was small (-0.92pp) versus a much larger average 10am→final move (3.51pp), so much of the day’s information arrived late. Where you see aggressive early support (as with Henfield), treat it differently from a gentle late squeeze.
6. Sample size discipline matters: one day can be noisy, but the 30-day pattern is not. If you are weighting market signals in your process, weight the rolling evidence more heavily than yesterday’s variance.
7. No filter-supplied qualifiers were available at generation time today, so there is no named short-list to anchor decisions; the Live Market needs to be read race-by-race using price behaviour and confirmation rather than pre-baked selections.