Football Musings

[wearthefoxhat]

I am wondering how much longer these models will hold up in the last part of the season, teams like Forest may play different as they will be hell bent to stay in top 4, teams just outside the top 4 will be hell bent on getting in. Teams at the bottom will be playing to stay up so may settle for a draw against stronger opp or look to defend a 1-0 lead against a teahm toward the bottom. Plus Liverpool with a big buffer may not field their strongest team against say Southampton, plus they have a lot of home games left so they may have a few weaker sides going out at home.

That's certainly something to factor in over last 20% of the season.

The way I'll likely approach it, is to take a view over a shorter time frame rather than a mid-long term view. ie: Last 4-6 home/away, rather than last 8 or the whole season stats. That way, the form cycle trends, and choice of players, will be easier to see and their influences,

The live market will take these factors into account too, especially the games of importance, so it's even more critical that you/we have an accurate xG model in place to make the comparisons. Could be a simple decision to reduce the stake % and focus on the summer leagues.

[wearthefoxhat]

I couldn't get it to perform as it should, GPT recommended for using in play and I looked at a 2-0 score line to see what the odds were for 2-2 and it gave a prob of 0.8% so I didn't continue with it.

The Bivariate model does stack up differently to the ZIP model. Just getting it to produce a formula in one cell was a herculian effort on my part. I tend to deal with in-play differently and prefer to use these models for pre-off comparisons.

Here's just half of the formula and how it looks at the lambda's and other stuff in the spreadsheet.

formula.png

After GPT produced something, I'd ping it over to DeekSeek, then ping it back to GPT. Let's just say, if it wasn't for this A.I malarkey, I'd have no scooby doo and be living in the dark ages. It's certainly helping an old boy like me close the gap and maybe, in some cases, overtake a few that have become complacent in their thinking.

[andy28]

I was getting tired with it as it was taking a long time to make progress, I am happy with that but I will take a day or two away from it and then will try again. I always say to the AI to dumb it down and explain it as if I were a 6yo. It does seem to help

[wearthefoxhat]

I was getting tired with it as it was taking a long time to make progress, I am happy with that but I will take a day or two away from it and then will try again. I always say to the AI to dumb it down and explain it as if I were a 6yo. It does seem to help

That reminds me of the character John Tuld (Jeremy Irons) in Margin Call. (must watch film)

JohnTuld.png

[wearthefoxhat]

One area, thinking ahead, is if there is consistent value identified pre-off, then a combination of acca's that multiply the "edge," could be a way forward. Small stakes, big wins and the bookies luv an acca....

Chris Forwood talks about it on his website, Sharp Betting, it's also scripted in his blog.


https://sharpbetting.co.uk/video/footba ... 00k-profit

CF2.png

[TupleVision]

Hey WtFH, I've recently been experiment with Bivariate Poisson, not sure if you've considered this but I'm getting better results than Zero Inflated so far.

Took a look at Bivariate Poisson last night and earlier this morning. Of course, I leaned on GPT and DeepSeek (free versions), GPT seemed a bit more user-friendly and picked up on my input errors. It was very patient, as I kept referring it to Bavarian Poisson...I'd never heard of it before until you mentioned it!



Why Biv.png


BiviaratePoisson.png


I'll run it alongside my Zippidy-doo-da poisson that seems to be holding up well....so far

Noice! I'm getting some great results with my in-play model using Bivariate I must admit. It adjusts a lot quicker to slight shifts in momentum and xG that you don't get with ZIP, personally I wouldn't use it for pre-match it really comes into it's own in-play. Naturally I used GPT for a lot of the heavy lifting too, although I've found GPT is better at the logic but coding is pretty poor so I use GitHub CoPilot for the actual python stuff.

I've pretty much abandoned my pre-match/halftime model now and purely focussing on in-play, the edge I've found is a lot better and I've adjusted so it will also give me goal probability which if you target the correct type of match before kick off the odds don't seem to adjust quickly enough after a goal sometimes on the over/under market allowing for momentum shifts, some juicy profits to be had there too.

[wearthefoxhat]

Hey WtFH, I've recently been experiment with Bivariate Poisson, not sure if you've considered this but I'm getting better results than Zero Inflated so far.

Took a look at Bivariate Poisson last night and earlier this morning. Of course, I leaned on GPT and DeepSeek (free versions), GPT seemed a bit more user-friendly and picked up on my input errors. It was very patient, as I kept referring it to Bavarian Poisson...I'd never heard of it before until you mentioned it!



Why Biv.png


BiviaratePoisson.png


I'll run it alongside my Zippidy-doo-da poisson that seems to be holding up well....so far

Noice! I'm getting some great results with my in-play model using Bivariate I must admit. It adjusts a lot quicker to slight shifts in momentum and xG that you don't get with ZIP, personally I wouldn't use it for pre-match it really comes into it's own in-play. Naturally I used GPT for a lot of the heavy lifting too, although I've found GPT is better at the logic but coding is pretty poor so I use GitHub CoPilot for the actual python stuff.

I've pretty much abandoned my pre-match/halftime model now and purely focussing on in-play, the edge I've found is a lot better and I've adjusted so it will also give me goal probability which if you target the correct type of match before kick off the odds don't seem to adjust quickly enough after a goal sometimes on the over/under market allowing for momentum shifts, some juicy profits to be had there too.

That's good to hear.

It's very much about finding your niche, pre-off or in-play. Reckon anyone with coding knowledge that can lean on A.I. should be okay moving forward.

[wearthefoxhat]

Was catching up with some you-tube stuff and a recommendation came across the ASP (assured soccer profits). I was a member of an old forum that shared excel and .pdf's and one of them was the original ASP.pdf. (96 pages).

Playing devils advocate, I asked GPT had it heard of it....

ASP 1.png

ASP 2.png

With my recent ZIP, xG metrics in-play endeavours, it would be fairly straight forward to find a way to trade value-driven games and automate it.

One thing I've noticed, is that this GPT model (free version) is second guessing a lot what I want to do next, heck, it might even know I'm typing this message into this forum....

[Kai]

One thing I've noticed, is that this GPT model (free version) is second guessing a lot what I want to do next, heck, it might even know I'm typing this message into this forum....

Its memory features shape how it responds so it tries to anticipate your needs. Some see that as bias and not personalization so they turn it off.

You can delete chats you don't want it to remember. But I'd suggest reading its memory at some point, it can be a bit uncanny seeing yourself through its prism.

It's very much about finding your niche, pre-off or in-play. Reckon anyone with coding knowledge that can lean on A.I. should be okay moving forward.

Think it's even more useful to those with zero previous coding experience/interest like myself, but plenty of market experience. It certainly removes a huge barrier to entry to some brand new possibilities, even of old ideas.

KUTGW with the thread

[andy28]

I looked at the MLS matches today and the penny may have dropped for me

Match 1: New England Game
Initially, I was skeptical about the MLS season due to low liquidity in the match odds (less than $20k), but I found a match with significantly higher liquidity. According to Guardian the expected goals for the New England match were 2.94. At 32 minutes, I saw the under market at 1.60, so I checked SofaScore, which indicated very low xG for both teams 0.1 for the home team and 0.00 for the away team. Possession was 30% home and 70% away.

I backed under for $20 at 1.60. By halftime, nothing notable happened the xG was still low (0.1 & 0.05). The odds had dropped to 1.3.

In the 52nd minute, the away team scored. I chatted with Copilot which suggested backing under again. The home side still had minimal xG, and the away side held possession but hadn't significantly improved their goal threat. So, I backed under again, considering it a safe bet. The home side had a second-half xG of 0.00, and the match ended 0-1. I could have greened up at around 1.20 but decided to stay in, resulting in a profit of $23.

Match 2: Similar setup
I came across another match that was 0-0 at halftime, with under odds at 1.30. I backed $20 on under. A goal was scored and the odds moved to 1.25. After assessing the game, I placed another bet. One team had zero goal creation, so I held until full-time, and it ended 1-0, yielding another nice profit.

In the second match, I ran the first-half stats through a Poisson distribution, which indicated a 98% chance of under 2.5 goals. Considering team talks, subs, and potential red cards, I estimated a 90% chance, making 1.30 odds look valuable. The game dynamics didn't show any attributes of an open game, so even with a goal I remained confident in my bet.


So is this the value talked about? Previously I would have gotten out after a goal or maybe in the first match at halftime, but with the aid of these stats it gave me confidence to stay in I did think greening at any point was bad value

Sorry for the long post but I am a wee bit chuffed with myself

[wearthefoxhat]

Its memory features shape how it responds so it tries to anticipate your needs. Some see that as bias and not personalization so they turn it off.

You can delete chats you don't want it to remember. But I'd suggest reading its memory at some point, it can be a bit uncanny seeing yourself through its prism.

Shaun mentioned the memory settings in a previous post. Looks okay, nothing out of kilter so I'll keep it switched on as I don't mind being prompted by questions and ideas moving forward. It's on the user to squeeze out the best with the right follow up questions and enquiries.

-----

Think it's even more useful to those with zero previous coding experience/interest like myself, but plenty of market experience. It certainly removes a huge barrier to entry to some brand new possibilities, even of old ideas.

KUTGW with the thread

That's very important, especially for someone like me. I haven't yet gone down that route, will likely crack the VBA stuff first.

My youngest son came out of university a while back with a top honours degree in computer science, but can't pin him down as he's currently working on a medical application to save lives, so I'll lean on A.I. for the time being.

[wearthefoxhat]

I looked at the MLS matches today and the penny may have dropped for me

Match 1: New England Game

Initially, I was skeptical about the MLS season due to low liquidity in the match odds (less than $20k), but I found a match with significantly higher liquidity. According to Guardian the expected goals for the New England match were 2.94. At 32 minutes, I saw the under market at 1.60, so I checked SofaScore, which indicated very low xG for both teams 0.1 for the home team and 0.00 for the away team. Possession was 30% home and 70% away.

I backed under for $20 at 1.60. By halftime, nothing notable happened the xG was still low (0.1 & 0.05). The odds had dropped to 1.3.

In the 52nd minute, the away team scored. I chatted with Copilot which suggested backing under again. The home side still had minimal xG, and the away side held possession but hadn't significantly improved their goal threat. So, I backed under again, considering it a safe bet. The home side had a second-half xG of 0.00, and the match ended 0-1. I could have greened up at around 1.20 but decided to stay in, resulting in a profit of $23.

So is this the value talked about? Previously I would have gotten out after a goal or maybe in the first match at halftime, but with the aid of these stats it gave me confidence to stay in I did think greening at any point was bad value

Sorry for the long post but I am a wee bit chuffed with myself

That's a very good approach, well done, and you should be chuffed.

Re-Value, ultimately it's "your value" trading decision, based on certain poisson metrics and A.I guidance, to come up with an in-play trade. I like the bit where you had a chat with co-pilot, sounds as though you're flying a plane and deciding which route to take to come into land.

[andy28]

I was just going off sofascore and then I found live streaming. Not being up to play with football like you guys I use Co pilot to explain formations and what the manager is wanting to acheive. When I got pictures in the 2nd game the formation was 4-2-3-1 at kick off and after the goal they went to 5-4-1 on defense which as co pilot said is they are parking the bus, with the home team having an Xg of 0.03 2nd half and .3 first there was little chance it was going to end up with 3 or more goals with 15 left.

Most football is play when I am stacking Z's and the A League has no Xg the MLS fits me perfectly 10.00am-6pm, so I can look for matches that fit this set up, I here Peter talking about a Racing setup so I see no reason why I cant have a football set up that suits me. Also I thought about the liquidity and as my bet are only $20 (10 pound) it shouldn't be a problem

[wearthefoxhat]

I was just going off sofascore and then I found live streaming. Not being up to play with football like you guys I use Co pilot to explain formations and what the manager is wanting to acheive. When I got pictures in the 2nd game the formation was 4-2-3-1 at kick off and after the goal they went to 5-4-1 on defense which as co pilot said is they are parking the bus, with the home team having an Xg of 0.03 2nd half and .3 first there was little chance it was going to end up with 3 or more goals with 15 left.

Most football is play when I am stacking Z's and the A League has no Xg the MLS fits me perfectly 10.00am-6pm, so I can look for matches that fit this set up, I here Peter talking about a Racing setup so I see no reason why I cant have a football set up that suits me. Also I thought about the liquidity and as my bet are only $20 (10 pound) it shouldn't be a problem

The formation of a team is a moving feast. Certain managers favour a set formation irrespective of the opposition, others will do their homework and look for weaknesses, be it down the left/right flank, over the top or down the middle. A red card or an enforced change (injury) can mess up a previous game plan, but a simple booking can do so as well. ie: Defender gets a yellow card early on, and now has to tread carefully and not make a rash challenge in fear of a second yellow.

If you're at the testing stage, working within your comfort zone for staking and a % of the bank is more important. Ideally, compounding from the profits going forward is best. Even then, I'd keep within comfort zone thresholds as it might skew your thinking whether or not to stick or twist when in a trade.

I've expanded all my excel sheets across all the european top leagues and i've got it the stage of a simple copy/paste of key data and everything populates for a pre-off check. Web scraping and other VBA stuff is next on the list.

[andy28]

I was just going off sofascore and then I found live streaming. Not being up to play with football like you guys I use Co pilot to explain formations and what the manager is wanting to acheive. When I got pictures in the 2nd game the formation was 4-2-3-1 at kick off and after the goal they went to 5-4-1 on defense which as co pilot said is they are parking the bus, with the home team having an Xg of 0.03 2nd half and .3 first there was little chance it was going to end up with 3 or more goals with 15 left.

Most football is play when I am stacking Z's and the A League has no Xg the MLS fits me perfectly 10.00am-6pm, so I can look for matches that fit this set up, I here Peter talking about a Racing setup so I see no reason why I cant have a football set up that suits me. Also I thought about the liquidity and as my bet are only $20 (10 pound) it shouldn't be a problem

The formation of a team is a moving feast. Certain managers favour a set formation irrespective of the opposition, others will do their homework and look for weaknesses, be it down the left/right flank, over the top or down the middle. A red card or an enforced change (injury) can mess up a previous game plan, but a simple booking can do so as well. ie: Defender gets a yellow card early on, and now has to tread carefully and not make a rash challenge in fear of a second yellow.

If you're at the testing stage, working within your comfort zone for staking and a % of the bank is more important. Ideally, compounding from the profits going forward is best. Even then, I'd keep within comfort zone thresholds as it might skew your thinking whether or not to stick or twist when in a trade.

I've expanded all my excel sheets across all the european top leagues and i've got it the stage of a simple copy/paste of key data and everything populates for a pre-off check. Web scraping and other VBA stuff is next on the list.

I didn't realise the importance of the formation, I was looking at Liverpool and their go to for EPL is 4-2-3-1 but in the CL or FA Cup it is 4-3-3 . So I ask Co pilot why and what are they trying to do, it makes for interesting reading.

I agree with all you said I did use bigger stakes (for me) than normal as I was going to trade out at Half Time or at some other point, but with the lack of goal creation chances I kept asking myself, Why? There was no reason to and it turned out that way.

I used AI to do a correlation of some of my stats to see what plays the biggest in goals so I just kept checking those as well as the game unfolded. But Co pilot did say "Correlation is not Causation" so I do keep that in mind.

What impressed me the most was I got the most I could from the game (profit wise) as I know there will be a game where two quick goals happen and in the blink of an eye it is 1-1, at that point anything could happen, but as I am making more from winning bets I will be more at ease to get out of those matches where it goes against me, old me would stay in and hope for the best.

All the best with the VBA stuff that sounds interesting