Trading Naked - using bots while asleep :)
-
Thanks for looking in. Bit of background about this project to set the scene. I work permanent nights on a Sunday, Monday, Tuesday and Wednesday, six though to six, generally getting home around 07:15 am and going to bed about 08:15 to be up and out by 16:45. This seriously reduces the time I have available for hands on trading, basically a Friday and Saturday being my only real options, however as you can imagine the fact that I am not home in the evenings, I have a lot of pressure to not spend Saturdays in my cave in front of a PC! (Although I took time off this week, so have more time to focus on trading projects.)
So, I have to use automation to do my trading for me, both for set and forget and general in out trades. The problem with automation is that while it is great for automatically placing bets and greening / redding etc, when it goes wrong it goes wrong it REALLY goes wrong! You can loose a lot of money very quickly as I have learnt the hard way. You need to spend a lot of time watching it execute (and not interfering) and seeing what it does, hundreds and hundreds of times on many markets, tweaking and then going again. This is what I have been up to for the last few weeks and thought I’d share my progress as it may be useful to those who are not yet using bots and for me to get advice from people who may have more experience than me.
The bot I am working on started life as a basic lay the field at 3. This requires 4 matches to get profit and is a pretty unreliable bet. Where the bot excels on this is that it will see profit really quickly in running and green - I am not greedy, I am only aiming for 10% of a point profit per run (essitially an in running scalp), even favourites at just under 2 will generally move in running to give a 10% shift , usually in the latter part of the race when the tussle starts for the line. Case in point, the 13:30 at Wincanton today (9th March), Stormy Flight was trading at 1.2 up until the last furlong or so when Windance came in from 2.5 to win (My bot had greened around 10 Furlongs in to the race, when Windance went to 3.65 for a bit from 3 - so the final furlong drama was not really an issue but just giving the example of how fast stuff changes in the last few seconds of a race).
I am hoping that I can post a weekly update on progress with the testing of this bot. I am going through last weeks data at the moment, early indicators show - and bear in mind the current version is only mechanics, it has no stop loss or countermeasures in it, basically a straight lay the field at three cancel all unmatched bets when greening (if it can green that it is!). One version has a limit of not running when the favourite is under 2.6 at 1 min before the off and the other version just goes in and runs no matter what (so it gets all the short favourites) - but both versions only lost 6 and 10 points respectively in the week and that includes the outage on Saturday at Betfair and a -8 point disaster at Ffos Las on Sunday.
They run on EVERY race every day - in reality this would not happen or be sensible, but I am encouraged that they have lost so little without any countermeasures in place. The other purpose of running them on every race is that I can generate loads of data, they are collecting in running actions alongside so I can see the positions/prices of other runners, book %, traded volumes and so on when things happen with the bot. Adds hours to the data analysis, but worth it!
Gone on a bit there, I will do an overview of my testing set up in another post in a bit as I think it may be useful to others.
-
Not been very vocal over the last few weeks but that does not been I have not been testing things. So an update, I was waiting to see what the happened up to mid May with the lay at 3 automation - I knew pretty much that it was not going to go too well, even filtering results by type and distance. Essentially, it would not work as a set and forget, it is difficult to get three under 3 consistently enough to make it worth doing. So, this is going to be moth balled for now.
However, what I did spot is that in races where the favourite goes off over 3.5 it could be worth backing the 1st, 2nd and 3rd favourites. I have been testing this for a couple of weeks now with a 1 point back stake per runner and it does seem to be interesting enough to take a deeper look. Early indicators seem to say Flat / AW only and 7 to 11 furlongs are the best bets. You don’t get many per day and the minimum return is 0.5 points to a max 3 point loss, but the strike rate is pretty good so it is mitigating loss.
Now that I’ve knocked the lay at 3 on the head, I’ll set up a couple of test scenarios, one flat backing set and forget, the other a trading version, to green in running and see what happens. You gotta love science LOL.
I’m also testing a couple of other automations / strategies - in particular an in running one that if the 2nd favourite is below 1.8 then lay the first and second - once again not many a day but it is too early to tell what’s what - partial matching is a problem though so this may not be viable Only time and testing will tell me I guess.
Another one that lays the LTO Winner who is a course winner more than twice (distance winner either never or not), and a few variations, but that is just a paper trading affair at the moment pulling from the software only. It seems interesting, my filter gets an Adam score of 3.5 which is encouraging.
-
@richard-coxon thanks bud
-
This is for Nick Allen ( @nick-allan ). As I said in the PM, this is untested but it uses relative odds condition to check if the selection is below 4th favourite (plus the other conditions), if so it will fire, if not it wont.
0_1680201329547_BTC Back CD Winner.baf
As I said I have not tested it, although I think it will do what you wanted. One note, I can NEVER remember when using relative odds if you need to have it as plus 1 Tick or Minus 1 Tick - logically I think it is plus 1 Tick - you'll see in the conditions. Anyway hope it helps, if not then PM me :)) There is bound to be a CD winner selection over the weekend I'll give it a whirl myself and see what happens.
-
@richard-coxon great to hear!
Yeah always good to keep a negative data set in testing if you can, then you can see if the trend continues.
-
So, just finished for the day, totaled the numbers and added everything for the week - Monday to Friday - for the one I have shortlisted, so this is a quick and dirty result .. a raw total of 3.3 points profit. I can see where I appear to be making the most losses, first where it matches on more than one selection and secondly 5 furlong races.
I am not going to remove 5 furlong races at the moment, sticking with them, why? Because there is a higher number of matches on these. So what I am going to do it say once there is one matched bet, cancel all unmatched bets. At the end of the day, I can filter these out in Excel, so may as well continue to collerct data.
I hasten to add - these results are in practice mode .. so are probably a bit kinder to me than in a live market
-
@richard-coxon a cautionary tale indeed, thank god you had notifications set on Betfair!
-
HAA HAA, a cautionary tale! I set up a new instance of Bet Angel on one of my test boxes this morning. Now I was in a bit of a rush as I was going for a hosptal appointment and wanted it to get going on the card for today as it would start before I got back. I was walking back to the car and started getting notifications on my phone from Betfair! I'd forgot to set it to run in practice mode .. luckily it had only lost half a point by the time I got back, but that could have been a real disaster! On the plus side, it went pretty well on live markets, every cloud has a silver lining I guess. But big note to self - CHECK SETTINGS TWICE then check them again!
-
@greenie Cheers, at the moment I am working out the logic for greening effectively using signals, it is working quite well. Today, for example, it fired on 16 races across the card - with no prior race type selection (I was going to do that but decided not to while Cheltenham is on) and successfully greened on 15 for a 1.8 point profit, loosing 0.7 points on a miler at Newcastle. I'm quite happy with the greening now, seems to successfully do it. My next mission is to work out rules for taking a red, I tinkered a bit last week but actually lost more points than if I left them, dunno what to do to be fair at the moment. Let them ride and take the hit, but that depends on strike rate and and maybe extending the green limit to over 10% of the net stake. Smells of gambling a bit to me LOL. But I am sure I can work it out, thinking of setting a counter on the odds of the selections and red if the odds are too short and the next nearest is further than x amount of ticks, dunno. Maybe it wont be an issue if my race / runner selection criteria is good at picking 3 legged horses I certainly used to manage that pretty well when trying to pick winners !
-
Great first few posts here - i'm eager to keep up to date with your progress. I'm in a similar position where work commitments mean i'm relying more on the Automation side and set and forget strategies than manual trading. I went through a stage of manually trading in running horses using BetAngel but it required alot of focus and time.
I'm currently experimenting with various strategies on BE Guardian myself.
Best of luck and I look forward to seeing future updates!
-
@richard-coxon have you tried hooking these up to a tv? just wondering if it could replace my chromebooks I use for streaming live sport to tv
-
@richard-coxon wow great price that too!
-
@martin I bought a new one just last week. I had an 'old' fanless one with Windows 10 version 1909 on it that was not upgrade-able to a later version of windows and the network adapter was a bit dodgy, so I needed (well wanted) to replace it. I got this https://www.geekom.co.uk/geekom-miniair-11-mini-pc which gives plenty of bang for the buck, the N5095 is pretty capable for what I want it to do, the fact that there is scope for physical upgrades of RAM and SSD is a bonus (a lot are soldered RAM / SSD so not upgrade-able). It is very quiet compared to my other mini PCs. If ever I needed to I can hook it up to two monitors, so perfectly usable as a standby desktop, it has Type C USB, so can bung a hub on if necessary and a LAN port, so can go straight into the router via a hub, so no need for WiFi and can make full use of the network adapter. The only thing I don't like is the power adapter in the version I got is a bit bulky, but can change that. I am going to get another in a few months. Geekom are a reliable well trusted brand too.
-
@richard-coxon always been curious on mini PCs but never had the courage to go for one, sound like they work well for this!
-
@richard-coxon I really didn't wanna know that
-
Change of plan, as it is Cheltenham this week I have decided not to start a fresh cycle on a tweaked bot this week. I'm going to leave everything as it is. Mainly as Cheltenham will have a big impact on the 'other' meetings so best wait for a 'usual' week.
-
As promised, a brief description of my testing set up. I am using Bet Angel, the Guardian software, once you get your head round it, is fantastic. The licence for Bet Angel is per login, so you can run it on as many computers as you like. This is great for my needs. I have 4 mini PCs - these are Intel Celeron based, with 4 to 8 Gb RAM and between 64Gb - 128 GB SSDs. Bet Angel does not require high spec PCs to run, just lots of bandwidth over the network. You can run multiple instances of Bet Angel on the same PC if you want to, however you end up bottlenecking the network adapter and as you want to be running the Guardian software at less than 100ms refresh rates, running over Wifi on the same PC is not great. Hence the mini PCs. They are cheap, around £120 - £200 each and once you have them set up you can use VNC to access them, so no need for keyboards, mice or monitors, they just run.
I access them through a web browser to configure the days races (Safari or Firefox have a VNC applet built in on the vnc:// protocol) Or you can use Chicken of the VNC on Mac or RealVNC on PC. I have a batch file on each of the mini PCs that runs at midnight each day that copies the log files across to my Onedrive (I have a folder set up for each machine) so I can then process the days results - well weeks results I don’t look at each day as it happens - it’s too tempting to fiddle after a bad single day. The reverse is true for updating the bot files, I export from my Mac (I use Parallels desktop and run Windows on my Mac) to Onedrive then import the bot files I am testing to each of the mini PCs.
I have chosen to run the various version for 2 weeks each to see how they are performing, this week is a decision week. What I mean by this is, they have all been running different versions , one does all, another only if the favourite is over 2.6, another cancels unmatched bets after a match in running and another has a rule that lays 2nd, 3rd. 4th, 5th if the favourite is under 2.6. So we will see the results and I’ll decide which one to run over the next two weeks across all the machines with various tweaks, so they will all be running the same bot, but with different settings.
The other change this coming test is race selection, I won’t be running the bots on Maidens, Novice, Beginners, Bumpers. Amateurs etc. The data so far indicates that theses are terrible! Most of the losses have come from theses types of races (but I anticipated this would be the case as in the past they have performed badly on previous automations) the odds movements are too mad and unpredictable.
I’ll post my results for the previous week in a few days as I will be downloading last week after midnight tonight (if it is quiet at work I may load a spreadsheet up for a quick peek).
All this may sound complicated and time consuming, but I can assure you, once you have everything set up its really straighforward. I get in in a morning after work, get a brew and can have all 4 machines set up for the day in 15 to 20 mins, then off to bed!