Pre-race L2B: gating a close-trade rule on whether the position is still open

Advanced automation available in Guardian - Chat with others and share files here.
Post Reply
User avatar
aljacko1
Posts: 14
Joined: Tue Oct 22, 2024 6:35 pm

Dear forum members,

I asked a version of this on 3 May in the venue volume filtering thread and it went unanswered. I now have data rather than a hypothetical, so I am asking it properly in its own thread.

Background: pre-race lay-to-back on the favourite, UK win markets only, £10 stakes, practice mode throughout. Entry lays at −15:00 with a 3-tick offset back and a 10-tick stop. A close-trade rule armed −00:30 to −00:00.

What I found in September: the closing rule had been type "Close Trade On Selection" rather than "...With Greening" since May. It placed an equal-stake opposite bet, which levels nothing — around 70% of markets settled at exactly £0.00 and the rest at a loss, because that is all an equal-stake close can produce. Greening fired zero times in over 1,700 trades. Rebuilt as v9.0 on 8 September with the greening type; £0.00 outcomes fell from 71% to 1%. The mean barely moved, which is what I expected — greening is a variance fix, not an edge.

The open question is about a condition I deleted in the same rebuild.

The old build had a condition on the closer: best back price > stored lay entry price − 6 ticks. I removed it as part of the rebuild. Looking at the logs afterwards:

- Under the old build: 8 stop events, 0 followed by a pre-off closure.
- Under v9.0: 31 stop events, 22 followed by a pre-off greening closure.

So that condition was doing real work. It was implementing what Bet Angel told me on 27 April — that the closing rule needs a condition ensuring it only triggers once the stop bet from the entry rule is no longer active. I did not realise that when I deleted it.

I genuinely do not know whether removing it helped or hurt. Greening after a stop books a small certain loss rather than running an unhedged position into the race, which may be the better outcome. 174 trades in, the distribution is tighter and the mean is unchanged. I am collecting to around 500 before deciding anything.

Question 1 — has anyone built a condition that reliably answers "is this position still open?"

Bet Angel support told me this week that the Close Trade Profit condition returns the selection's trading profit (divided by current price when greening is ticked), so a greened position and an open one both return non-zero and it cannot separate them. They pointed me at Net stake instead, with the caveat that it will not be exactly zero after a greening close because the greening uses a small stake to spread the profit. I have asked what threshold would separate the two states and I am waiting on that.

Has anyone actually built this? Either a stop-state gate, or ShaunWhite's suggestion from May — firing the pre-off greening bet and setting the unmatched portion to TakeSP, which I still have not implemented:

"If you don't want to carry liability through to in-play then you could fire your preoff greening bet and then set it (or the unmatched portion) to TakeSp."

Question 2 — trigger count on the closer, 5 or 10?

I have two sources that disagree. jamesedwards, in the Greening Close placement speed thread in May: "I would set it to green 10 times every 3 seconds between 30 secs and 0 secs out." Bet Angel support, on the same question: trigger the closer 3–5 times, rearming every few seconds, on the grounds of preventing a machine-gunning escalation I had in April.

v9.0 runs 5 triggers at 3 seconds, which covers roughly 9–15 seconds of a 30-second window. Bet Angel have since confirmed that each new green-up cancels prior unmatched bets, so triggers replace rather than stack — which seems to weaken the escalation argument for keeping it low. Against that, ShaunWhite's point from the same thread: "Chasing a price with mulpile hedges in an illiquid market will erode your edge more than you think."

Has anyone measured the trade-off between fill rate and price concession on repeated greening attempts? I have not, and I have not seen it measured anywhere.

Happy to post the per-day figures if they are useful. Everything above is practice mode.
Post Reply

Return to “Bet Angel - Automation”