Spent two days upgrading and back testing model... Starting new 10 day trial...

**prolexus** · 02-11-24, 12:38 PM

NCAA Basketball takes too much time given I have the other numerous models...

**Neek762** · 02-11-24, 01:48 PM

Curious to which sharp book you both are using as the source of truth when comparing your lines

**usma1992** · 02-11-24, 03:07 PM

Prolexus...

My models are fully automated... so it really doesn't matter much how many teams there are either way. Teams change year to year so I need to make sure they match up but in all honestly it takes me 1 minute to run pro NBA and a minute twenty to run college Basketball... what do I care how many teams there are in the system. The more the merrier.

Neek:
I use Betonline to get my lines... they seem to be as accurate as the rest of them. However, on the programming side if I decide to switch to another site... I have to work on the name change piece.

If you don't know... the names are the biggest pains of all of it. Some sites say Florida State... some FSU... some Florida St some Florida St. .... with a period. As a programmer this becomes a dilemma of how to match up the right statistics with the right team. In addition, when I download the results... I want the results to populate the right teams.

Dave

**Neek762** · 02-11-24, 03:20 PM

Yeah you'd have to figure a way to scrape the matching rotation numbers so that everythings in sync... i feel you though because there's already a bunch of double checking as it is placing wagers on the correct side lol

**usma1992** · 02-11-24, 03:26 PM

Nothing today... 48-31... almost 61%.... that doesn't include the previous bets before Jan 2.

I appreciate any reasonable feedback.

Dave

**bjb7223** · 02-11-24, 05:24 PM

Originally posted by usma1992

Nothing today... 48-31... almost 61%.... that doesn't include the previous bets before Jan 2.

I appreciate any reasonable feedback.

Dave

stick to college hoops and forget about the NBA

**usma1992** · 02-12-24, 07:12 AM

12:00 AM	KANSASCITY	22	SANFRANCISCO	21		43	-2	47.5		2Close	Under		0	0
12:00 AM	SANFRANCISCO	20	KANSASCITY	22		42	2	47.5		2Close	Under		0	0

**usma1992** · 02-12-24, 08:27 AM

I was close on SuperBowl lol...

No college bets tonight... One NBA... still trying.

Dave

**Neek762** · 02-12-24, 02:41 PM

Hm ... I wonder how much the time of day you're running your model is making an impact on picks. There's been a lot of value on the lines so far since open

**usma1992** · 02-12-24, 03:31 PM

Not sure I know what you mean. Not many picks lately is Vegas catching up to me.

**Neek762** · 02-12-24, 04:30 PM

Because there are point and total swings throughout the day. Total can open at 240 so your model will label it a no play, but it can swing 5 points in either direction by tip, so there still can be a play depending on how you're using your number. Not to mention injury news. That's the hardest thing about originating is the constant maintenance and adapting to the market

**prolexus** · 02-12-24, 04:44 PM

The best way to model is versus the Opening Line.

The Opening Line is Static and consistent. Throughout the day, lines are moving and may vary from site to site leaving uncertainty when determining plays.

Consistency is a Prime KEY ingredient when utilizing mathematical models.

**Neek762** · 02-12-24, 05:24 PM

That's why I was curious. I'm in CA and I notice he always releases picks around 8-9-10am. Not sure when the data / lines were scraped though. But plenty of bettors are getting down the night before. There can be a huge swing in value between the time the lines are released, by the time Dave is modeling, and by the time he posts the plays. So there can potentially be non-plays via the model that became non plays due to that time decay

**EasyCover** · 02-12-24, 07:10 PM

Appreciate you and you transparency.

**usma1992** · 02-13-24, 08:34 AM

Lost on NBA again... just one game... upgraded it again... Not giving up. I want all 4 of my systems to work. There is another NBA game tonight and it is as tight as it has ever been. Here is college. I am not sure what to do about the line movements. My system is extremely tight so when lines change the bets do too... I don't have time to rerun it over and over.

Time

AWAY TEAM

PROJ AWAY SCORE

HOME TEAM

PROJ HOME SCORE

TOTAL

VEGAS SPREAD

VEGAS O/U

SPREAD BET

O/U BET

Actual Away

Actual Home

8:30 PM

TENNESSEE ST

77

TENNESSEE TECH

65

142

4

136.5

Away Minus

Too Close

0

**RM Logic** · 02-13-24, 09:35 AM

The NBA is just a bad product and is too hard to handicap because of the BS variables that come into play that no other sport has. First you never know who is going to play or how much time players are going to play. There are games players and coaches could care less who wins. Then you have the nonsense officiating where nothing is called against certain players and refs make up calls all the time. When you have 10 incorrect calls in the last 2 minutes of a game (like happened a couple of weeks ago) you have to wonder just what is a legit game. College refs despise the NBA because it is not officiated with the goal of calling the game correctly.. You lost the under in the Philly game when there was 10 points scored in the last 50 seconds with zero chance of the team behind winning the game. It was nonsense. You dont see that in any other sport.

**usma1992** · 02-13-24, 09:54 AM

RM Logic... I hear you but I still feel like I should be able to overcome those inconsistencies with enough data and enough analysis. Maybe not, I have obviously not be able to up to this point. I am trending around 45% in my predictions which is obviously losing.

My current NBA algorithms are back tested to an 85% hit rate. I know that is looking backwards and not forwards... but I am trying to create specific algorithms. It is only picking games now that have a real shot at winning. I am starting new trial today. I only bet 10 a game on NBA.

I will run this for the next two weeks to see if... I can get it to work.

RM Logic... I am confused about one thing.

There are a few threads that get multiple views yet... their picks aren't winning consistently or at all. They fully admit they are upside down yet people continue to view them.

I feel like I am one of the few people that have produced a model that is actually winning... consistently. Am I wrong? Is there another thread out there that I should be paying attention to.

Dave

**Yensel** · 02-13-24, 12:26 PM

I registered for the sole purpose of thanking you and some other posters. So, here, in my first, Thank you for sharing, Dave!

**jamesrav** · 02-13-24, 12:32 PM

Originally posted by RM Logic

The NBA is just a bad product and is too hard to handicap because of the BS variables that come into play that no other sport has. First you never know who is going to play or how much time players are going to play. There are games players and coaches could care less who wins. Then you have the nonsense officiating where nothing is called against certain players and refs make up calls all the time. When you have 10 incorrect calls in the last 2 minutes of a game (like happened a couple of weeks ago) you have to wonder just what is a legit game. College refs despise the NBA because it is not officiated with the goal of calling the game correctly.. You lost the under in the Philly game when there was 10 points scored in the last 50 seconds with zero chance of the team behind winning the game. It was nonsense. You dont see that in any other sport.

but that should work out (equalize) in the long run. 10 points scored in the last 50 seconds to lose, will some day be the opposite, and you win. It's no different than losing (or winning) a bunch of horse races by a nose, neck, or head. If you play long enough it will even out. However, in looking at a lot of data on races like this, I found that heavy favorites 'somehow' did manage to win far more of these close races than could be expected by chance. Could it be that the 'best' horses simply have that little extra that allow them to win these close races more often than not?

**Neek762** · 02-13-24, 01:27 PM

Wouldn't that depend on type of race it's in? Sometimes they'll be running along side some donkeys or horses making their first start or first run in years so the favoritism is a little skewed. But when it comes to the higher grades I think it's just always the perfect storm for them to get the best of the others. Track, trainer, rider, postion, all that. So still tricky

**usma1992** · 02-13-24, 08:50 PM

Yensel... I appreciate it. It is a ton of work and this is by no means by full time job. I would still be doing it... but by posting it helps me legitimize it in my mind but I also appreciate comments to keep going.

I just spent the last 3 hours on the NBA Away+ bets. I was only looking at a singular bet in the NBA. Neverending.

Dave

**jamesrav** · 02-14-24, 03:36 AM

Originally posted by Neek762

Wouldn't that depend on type of race it's in? Sometimes they'll be running along side some donkeys or horses making their first start or first run in years so the favoritism is a little skewed. But when it comes to the higher grades I think it's just always the perfect storm for them to get the best of the others. Track, trainer, rider, postion, all that. So still tricky

in the last 10 yards of a race, when two horses are side by side and it comes down to the 'bob of the head' in some cases, there's no reason either horse should be chosen to win. Yet the heavy favorites were winning 60%, and the sample size was quite large. Could still be 'by chance' - that can never be ruled out - but maybe the better horses (generally with better jockeys) just want it a bit more. Back to the NBA: it being more difficult to beat compared to NCAA makes sense, fewer teams, lots of past history between teams. Trying to beat NBA, NFL, or Premier League Soccer seems almost impossible to me. I wonder where Billy Walters made most of his money, in the interview it seemed like he was betting mostly on College.

**usma1992** · 02-14-24, 09:48 AM

At a loss of what to do at this point... I am not concerned that I lost last nights bet. Tenn St. had 17-18 turnovers and didn't shoot well at all. I will accept that loss. I was happy with NBA for the first time. Sacramento was down by 1 with 24 seconds left. I think it was the right bet... got a push.

Here is the problem. It is now producing very few bets. It is like Vegas lines have finally caught up to me. There are no bets today for college or NBA. I expect it in NBA... not in college.

Do I just assume that my season is from mid November to Valentine's Day and end it there? Is the end of the season not going to be as lucrative as the numbers finally normalize?

Should I loosen up the program?

Dave
THE TEACH

**Sund** · 02-14-24, 10:33 AM

Continue to run it & see how it plays out ! A person has to choose whether to bet or not ! It’s by choice!

**Yensel** · 02-14-24, 10:43 AM

Well, please keep going. Love action, but if nothing qualifies, I thank you for your time. Understand how you could be frustrated after what you have put into the process.

**Neek762** · 02-14-24, 12:03 PM

I honestly think you are tweaking the model too much after each day. There will always always always be variance and I think you are trying to eliminate variance, which is eliminating picks. I think whatever you were doing back when you were producing 5-10 picks was working the best. From there, it seemed like you kept changing things, and as you kept changing things, the amount of picks steadily decreased. It will be hard to make anything of all of the work you've done if it's producing one play or less a day! I think you simplify a little bit and record all of the wagers and results and see what the results are around 500 picks before you make any changes.

**usma1992** · 02-14-24, 06:50 PM

I know these picks are last minute and there is a reason for that. I will explain in my next post. I might stay away from Uconn... its a big #.

Time	AWAY TEAM	PROJ AWAY SCORE	HOME TEAM	PROJ HOME SCORE	TOTAL	VEGAS SPREAD	VEGAS O/U	SPREAD BET	O/U BET	Actual Away	Actual Home
7:00 PM	UNC-GREENSBORO	79	THE CITADEL	67	146	4.5	136.5	Away Minus	Too Close	0	0
9:00 PM	CONNECTICUT	92	DEPAUL	55	147	24	141.5	Away Minus	Too Close	0	0
9:00 PM	TENNESSEE	82	ARKANSAS	65	147	8.5	151.5	Away Minus	Too Close	0	0

**usma1992** · 02-14-24, 07:10 PM

I appreciate everyone's input. Neek ... to your point. I really have barely touched it since Jan 2. The only thing I have done is look at individual games that roll in that I am not close on. Then I try to see what I missed and make minor swing adjustments. It is true that my tweaks may have eliminated future wagers that could have been winners. But at this point it is north of 60% and that counts predicted and past data. However, last Saturday it predicted one game out of 100. That is unacceptable. Yes, it won but still it is unacceptable.

So here is what I decided to do today. It might get difficult to follow. I noticed in college FOOTBALL that my projected scores without filters got very strong 2/3rds of the way through the season. The last 1/3 of the season I crushed it. My numbers were on.

So I decided to see what this season would look like without filters. I removed all the filters except for a small point filter to eliminate extremely close wins or losses. If I go back to Nov 15th... when I started... almost all categories of bets were dead on 50%. Some higher... some lower. So again, I am just looking at my projected totals with no significant changes.

So I then decided to look just at the last week and the numbers were excellent. So then I went back two weeks to Feb 1st and the numbers were still north of 63%. I then went back another week then another week. around 6 to 7 weeks back right around Jan 1st my numbers for unfiltered bets went from 59% to 56%... then continued to go down from there. The further I went back the further the more my winning percentage decreases.

So here is my conclusion... it may be wrong but I came to a simllar conclusion in college football. My system needs to be dynamic. For the first two thirds of the season until Jan 21-25... I need to have projections plus filters and algorithms that work.

Once I hit critical mass of games tracked for the season around 2500... I can then also rely on my projected totals and spreads as additional bets.

Therefore, on a daily basis I will be producing a reasonable amount of bets because I will produce the filtered and unfiltered at the same time. This is my plan moving forward. If it doesn't work I will relook but producing no bets or just one bet is not going to work.

Now because I save my models every fews days to google drive. I can always go a revisit previously algorithms and test to see which ones worked the best. Maybe as Neek stated my model Jan 1st is better than it is now. I did have two weekends that went 19-1 and 13-1 but then I followed that up with 1-5 and 1-0.

I also have to believe that maybe Vegas refinds its lines toward the end of the season and it is hard to get an edge.

Who knows... trying.

Dave
THE TEACH

**Neek762** · 02-14-24, 10:25 PM

Hm that's another thing that's considered - how far back you're going when compiling data for a game. I haven't done NCAAB, but my NBA model only used the last 5 games. There were times sides would go undefeated but the totals were off. There were times only a couple wides were correct but the totals were spot on. I woulld rarely change anything, i'd just look for patterns and bet accordingly. But when you brought how the farther back you go the closer to 50% you get, I'd have to imagine it's because how differently teams play through the season. Especially in NCAAB they have some dog shit matchups lol

**usma1992** · 02-15-24, 07:13 AM

3-0 Last Night on Unfiltered Bets... Over Tonight is Filtered so I would wager more on that bet... I'll come up with a system to identify filtered and unfiltered bets.

Time	AWAY TEAM	PROJ AWAY SCORE	HOME TEAM	PROJ HOME SCORE	TOTAL	VEGAS SPREAD	VEGAS O/U	SPREAD BET	O/U BET	Actual Away	Actual Home
7:00 PM	SMU	81	TULANE	71	153	2.5	158.5	Away Minus	Too Close	0	0
7:00 PM	N CAROLINA A&T	61	UNC-WILMINGTON	89	150	-21.5	141.5	Too Close	Over	0	0
7:00 PM	COLL OF CHARLESTON	82	NORTHEASTERN	69	151	4	153.0	Away Minus	Too Close	0	0
8:30 PM	W ILLINOIS	76	TENNESSEE TECH	64	140	3	127.0	Away Minus	Too Close	0	0
9:00 PM	LONG BEACH ST	68	CAL DAVIS	73	141	-3.5	154.0	Too Close	Under	0	0

**usma1992** · 02-15-24, 01:09 PM

So I did some soul searching about my program and the last few weeks and months last night. I reviewed this entire thread... and I realized that I am producing less bets now. However, many of the bets I was producing the last 6 weeks were ML bets. The payout on those bets weren't great and if you parlayed them... and lost one... you were out.

The ML venture really wasn't that lucrative. I tightened that system up but of course not it isn't generating many picks.

By adding the feature of my filtered bets.... which are the bets I have been generating from the beginning using complex algorithms to my now unfiltered bets which are solely my projected numbers... I am now generating enough real bets with hopefully a reasonable return.

For instance, last night... all three bets were unfiltered. I had no bets from my filtered system.

Tonight... there are five bets... only one is filtered. The OVER bet on UNC Wilmington. Let's see where tonight takes us.

So again, my lesson from college football... is to only bet filtered games until I have enough data. I am determining that to be about 2/3rd through the season.
51-31... after last night... 62%.
Dave

**Yensel** · 02-15-24, 02:36 PM

Thanks, Dave. Played all five of your picks this morning and may add some more OVER on A&T/Wilm. Also, really appreciate that you usually release picks early. Missed last night's rare late ones, but no worries. Any time that is good for you is great! Thank you very much for your continuing efforts.

**usma1992** · 02-15-24, 04:10 PM

Stay around 3% of roll per pick... anything can happen and I don't want you to lose your entire bankroll. Yesterday, late submittal was only because I had to spend 3-4 hours revamping... and adding unfiltered bets.

I was just glad that the unfiltered bets came through in flying colors. None of the games were even close. I am working to see if I can get NBA up to speed.

My new generated picks will tell you if they are filtered or not. Ideally, if the model picks a filtered pick... it should be bet heavier. Maybe, 4.5% of your bankroll.

I was wrong 51-32... forgot to include a loss... 61.4%.

Dave

**usma1992** · 02-15-24, 04:37 PM

So data from the season so far...

Nov 18-Feb 13 (Filtered)

152-45 (77.16%)

Jan 25-Feb 13 2500 games Tracked (Unfiltered)

47-23 (67.14%)

Total

197-68 (74.3%)

Data from Jan 2 forward is predicted data. But I am feeling fairly comfortable with my numbers. This does not include the 3 wins from last night.

Anyway, those numbers are big numbers... unrealistic... probably... but I got a lot of leeway to give and still be profitable.

Good Luck and Good Gambling
Dave

**Yensel** · 02-15-24, 05:46 PM

Perfecto! Gracias!