Practical advice of stop losses for Dutching

[Gcampb]

Hi all

Hope you dont mind the post.
I am wondering what the most practical approach is here using betangel. I have a pretty solid dutching strategy that is breakeven . What turns it profitable is minimising full loss of stakes and Its hit and miss with regards to my sloppy manual exectution and quite frustrating really

I am now looking to approach it differently and instead of manually trading out in play, I want to set stop losses at a certain position for each selection, so that if they drift I automatically close the position...

Just looking through Peter's videos in the academy and the user guide. I was hoping for some practical advice. Am I correct in saying that the global settings would. not trigger if im using the dutching tool? It was suggested that I use offset bet at the % I wanted - but I dont think I can use this global setting in conjunction with the dutching screen? (I didnt actually realise you could set the offset bet to act as a stop loss either, I aways assumed if I was backing, the offset could only be set at a lower price than entry)

If I were to use the ladder screen and the right click stop loss function, am I right in thinking that if these are set pre off, they do not persist into the off? So I would have to set these in play?

Also with regard to the stop loss function, the guide is good, but I want to understand how this works live. So the stop loss is set, if the price moves to stop loss, betangel will trigger and calculate and set a lay bet. Is this done at the best available price or at the stop loss price and how does it work if an order is not filled?

Appreciate any advice - posted here in the newbies thread and did try to find some old posts, but couldn't quite see the answers I was looking for - thanks all.

[rosecruz]

Perhaps learn to read races in-play... or you could sign up to TPD and start collecting live data, which you can then cross-reference with your model's selections so you can eventually automate exit points based on live data values. Some races favour front-runners, some favour horses in the middle of the pack, and in some, position makes no difference. If you study that, you might be able to make quick decisions in-play as soon as the pack forms.

Yesterday evening we saw a few races where fancied favourites ridden by Rossa Ryan and Oisin Murphy sat at the back of the pack. Look at your data and see what happened. This doesn't mean it will always play out the same way, BUT if you know these sorts of things, you can avoid certain situations and seek out more favourable ones.

I still get it wrong more often than I get it right, but the further I go down these rabbit holes, the more selective I become — and at least when I do get involved, I have a reason for it (whether I'm right or wrong). Dutching selections just because the model says so, and then trying to find the exact exit point in-play without reading the race or having live data... I don't know, it just seems very difficult.

[Gcampb]

Thanks for the reply.

The selection process is fairly robust so my strike rate is stable, I know how many races I should be finding a winner in and it has translated from multiyear backtests into live markets, but I understand what you are saying about live data. I have used tpd data before, years ago mind you when it wasn't uncommon for a tracker to be on the wrong horse At the moment, I have the video/commentary but I use the in-play trader window to help make decisions - I just find its the easiest/quickest way to interpret the market - especially in the closing stages of a race.

I went ahead and started using stop losses, I tried various % bands for drift however I was finding that this was clipping winners from the dutch far too often, even when placed up at extreme 250% drift from entry price and placed at different stages of the race, so I have settled with an even more extreme band which simply serves the purpose of limiting catastrophic full stakes losses in loosing Dutches. I need to monitor over a bigger sample to understand what the winner clip rate is at this high band, but provided it's a low enough rate, it should provide a baseline positive roi for the strategy.

I was reading more in the user guide, just trying to see if there is any obvious functionality im not utilising properly yet, but haven't found anything thats clearly going to help. I think the next steps would be to look more into servants and guardian - which is one aspect of betangel I haven't really dipped my toes into yet.

[rosecruz]

I've had a similar experience. I tried to find the right automatic exit point by looking at historical in-play lows for my model's top pick, and the tests consistently came back showing that holding until the end of the race produced the better ROI.

No threshold I tested beat letting it run.

I have also tested dutching top 2 and top 3, and again my model performs far better when only backing the top 1 ranked selection.

That's why I think the issue is model selection and not trade management - you might be trying to solve an unsolvable problem.

If the model says let it run for best ROI for top pick, what are the chances of finding a best exit place for more than 1 horse in the same race?

[rosecruz]

You might find this of interest too https://youtu.be/eXtPaAjViRE?t=292

[Gcampb]

I've had a similar experience. I tried to find the right automatic exit point by looking at historical in-play lows for my model's top pick, and the tests consistently came back showing that holding until the end of the race produced the better ROI.

No threshold I tested beat letting it run.

I have also tested dutching top 2 and top 3, and again my model performs far better when only backing the top 1 ranked selection.

That's why I think the issue is model selection and not trade management - you might be trying to solve an unsolvable problem.

If the model says let it run for best ROI for top pick, what are the chances of finding a best exit place for more than 1 horse in the same race?

The testing I’ve done on the selection side has been quite extensive. I provide selections for a parimutuel syndicate, so I have a good understanding of where the models strike rates and profitability are strongest. I agree that there are narrower, profitable ways to use the rankings but my objective is slightly different. I am trying to build a higher-frequency, lower-drawdown strategy that can support a much larger volume of stakes. The models do produce profitable narrow subsets, but they also produce strong winner containment across the top portion of the rankings. I am trying to exploit that broader probability concentration by dutching a compact group rather than relying on one horse.

The raw Dutch is only slightly slightly positive before execution improvements. The attraction is that the strike rate is high and the losing runs are relatively short. If I can reduce the average damage on losing races without repeatedly cutting the eventual winner, then even modest execution improvements can materially change the economics.

That is also why I am not really looking for one universal automatic exit point based only on historical in-play lows. I have already found that simple drift stops can cut too many winners, even when they are set very wide. The problem is that a temporary price move does not necessarily mean the runner is beaten.

Ive had a few days to get stuck into it and taken the time to explore guardian and servants. I still want to retain manual control in play, but I've managed to build out some specific automations and servants to do some automated work with stop losses

Screenshot 2026-07-14 083118.png

The idea is it detects the dutch bets and arms the catastrophic stop losses in play, during the race, i can choose to protect individual runners which reduces the stop losses further - this is to help protect horses that initially steam well below entry price. In the early stages of the race, the stop losses can be hit but the price needs to be maintained for a period of time before they get triggered which should help prevent them being triggered by a sudden spike that gets reversed. Since I am actively watching the race, I then trigger the 'late stage' rules, which is effectively the point in the race when the contenders start to appear and the race is in its closing stages, this will tighten the existing stop losses and rules but will arm new triggers for horses that trade under 2.0 and apply new stop losses, I need to test this further but the basic concept is to allow any live legs to go on to win, but if they reverse from under 2.0 past a certain point, say 4 for example that's when the final stop loss is triggered.

The big pain point I have been having is that the basic stop losses are killing too many winning legs when its set even at a significant drift point, the same with trailing stop losses.. I think you are correct in what you were saying regarding the price dynamics in different stages of the race so im hoping this solution will give me enough of a balance between automation & manual control. The overarching goal is simply to minimise full stake losses without limiting wins.