Why Win Selections Beat Each-Way in Greyhound Racing

30 May 2026racing-analysis · each-way · data

If you've ever looked at a greyhound race in a high-street shop, you'll have seen the choice between "win or each-way." Each-way looks like the safer choice — half the allocation on the win, half on the place — and beginners gravitate toward it for exactly that reason. The maths, however, doesn't agree. This article explains why TrapStats only flags win selections for greyhound racing, what the in-house data showed, and how each-way is priced against you when there are only six runners.

All performance figures here are simulated/paper results; past performance does not guarantee future results. This is research and education, not betting advice.

The structure of an each-way selection

An each-way (EW) selection is two allocations in one: one half on the win, one half on the place. UK greyhound place terms are typically:

  • 1/4 the price on the place leg.
  • Place positions: usually 2nd in a 6-runner field (some races may include 3rd).

So a 4-unit EW at price 6.0 means:

  • 2 units win at 6.0 → returns 12 units if it wins (10 units profit).
  • 2 units place at 6.0 × 1/4 = 1.5 → returns 3 units if placed (1 unit profit).

If the dog wins, you collect both legs: 12 + 3 = 15 units total return on a 4-unit allocation. If the dog places only: 0 from the win leg, 3 from the place. Net loss 1 unit. If the dog finishes 3rd or worse: total loss 4 units.

The place market is much more efficient

Here's the catch. The place market in UK greyhound racing is much tighter than the win market, in two senses:

  1. The market's margin is concentrated in the place leg. Place prices are derived from win prices via 1/4-the-price, and on short prices (e.g. an evens favourite) that gives place at 1.25 — paying only 0.25 units profit per unit allocated on the place half.
  2. Place outcomes are much more predictable than win outcomes. The favourite reaching the place is a near-certainty (>60%) in most races, so the implied place probability is heavily concentrated on a small number of dogs. The market knows this.

Combine these and the place leg is genuinely hard to beat. A model that's profitable on the win side might break even or lose on the place side because the market's margin is bigger there.

What the TrapStats backtest showed

We ran a 30-day out-of-sample window (after the April 2026 retrain, before the model had ever seen the test data) and measured simulated ROI for three selection types:

| Selection type | Threshold | ROI | |---|---|---| | Win | ev_win > 0.10 | +19% | | Place | ev_place > 0.10 | −11% | | Each-way | meets both above | −17% |

That last number deserves a moment. Each-way combines the profitable win signal with the loss-making place signal — and the place leg's losses are big enough to drag the combined result below zero. (These are paper figures; past performance does not guarantee future results.)

This isn't a quirk of our model. The same pattern shows up across most UK greyhound racing-analysis backtests: win signals of 10–25% become place signals of 0% to −15%. Each-way as a structure is engineered for the market's profit margin, not yours.

When does each-way make sense?

In horse racing — particularly with bigger fields (8+ runners and 1/5 the price on three places) — each-way can be a sensible value play. The longer place terms shift the maths.

In UK greyhound racing, where:

  • Fields are nearly always 6 runners.
  • Place is paid only for finishing 2nd (sometimes 3rd in special races).
  • Place terms are typically 1/4 the price.

The structure doesn't favour the analyst. The market's effective margin on the place leg is enough to wipe out a good win signal.

What TrapStats actually flags

Every Denis pick is win-only. The recommended selection field on every prediction is recommended_bet: "win" — no place option, no EW. The reasoning panel on the /denis page makes this explicit on every pick.

If you want a safer-feeling exposure, the principled alternatives are:

  • Smaller allocation on the win selection. Halve your allocation and you've replicated the "feels safer" intuition without paying the place-market tax.
  • Skip the race entirely. Not all races are worth flagging. Our ev_implausible skip counter has been telling us this loudly — most candidates each tick are rejected.
  • Flag only at longer prices. The 4-week analysis showed that selections at forecast price ≥3.5 won 52u while <3.5 lost 94u in simulation. That's exactly why we deployed the denis_safety_min_price=3.5 floor.

The honest answer to "but win-only feels riskier" is: it isn't — it's just less padded. The padding costs you more than it saves.

TL;DR

Each-way in UK greyhound racing is a structure that hides the place market's market-favourable maths inside the appearance of insurance. The win side carries the value; the place side carries the leak. We ran the analysis and the place leg cost 11pp of simulated ROI, dragging EW below zero entirely. Take the win selection or skip the race.

← Back to blog