Started a computer system almost 7 years ago... shelved it... but I'm back with app

**Orbison** · 09-22-21, 09:54 PM

hi Dave, many thanks for posting this. will be interesting to see how you do this week and going forward.

have you released the app yet and is it available in the google play store? if not, do you have a site where it can be downloaded from?

also, will you be posting your app's plays for NFL this weekend? sorry for all the questions.

**gauchojake** · 09-22-21, 10:02 PM

Oh boy

**KVB** · 09-22-21, 10:15 PM

Interesting work here.

Good Luck.

**Buckandadime** · 09-22-21, 11:11 PM

For years I've been telling posters, " Dave's not here, man."
I'll be damned if you don't pop up!!!
GL Dave..

**jjgold** · 09-23-21, 05:18 AM

Too early

Too random

**usma1992** · 09-23-21, 05:49 AM

I will post link to app... not on play store yet

Honestly, trying to fine tune things for a year old two but it has been tested on back data. All I ask is feedback on how I can make it better. JJ I agree week 4 in college is early because of early matches. FBS vs. FCS... etc.

Two things happened this weekend that gave me hope. It picked NE Jets dead on 25-6... which is a very random score. And it picked Fresno St. ML to beat UCLA.

I will post NFL tonight and give you link, if u want to look at it. Only for droids right now.

I'm a Comp Science and Math teacher in high school so I got to get ready for class lol. But my degree is EE from West Point. All my fellow teachers are now using my picks. Apple users I just send spreadsheet.

GL... It has Marshall ML tonight and Carolina for NFL.

I'm only trying to hit 57% or greater.

Dave

**jjgold** · 09-23-21, 06:26 AM

keep us updated
This is much better than trying to handicap games because that has not worked in over 70 years for anyone

**Tanko** · 09-23-21, 09:14 AM

Good luck Dave.
Hopefully, this works out.

What is the general basis for making the predictions?
Yards per play statistics, combination of of other systems (SP+, FPI, Massey...), opening line adjustmenting, etc..????

Also, how well did it perform when tested against a data set that was not used in the generation of the model.

**Tanko** · 09-23-21, 09:33 AM

How is the system making picks? For example, the Miami (OH) vs Army game. The score prediction is 16-46 Army with a spread of Army -8.5.
This would seem to be an obvious selection yet the "Spread Bet" column says "Too Close".

I ask because I'm leaning towards Army to cover this week.
Why the "Too Close" designation?

Also, I'm not taking shots at your system. Just trying to understand it.
I genuinely hope this works well.

**marcoloco** · 09-23-21, 11:06 AM

interesting

**ChuckyTheGoat** · 09-23-21, 11:14 AM

Originally posted by Buckandadime

For years I've been telling posters, " Dave's not here, man."
I'll be damned if you don't pop up!!!
GL Dave..

I'd like to tell u about "The Daves I Know"

**RudyRuetigger** · 09-23-21, 11:16 AM

weird thread to say the least

i dont have enough time to argue

**sweethook** · 09-23-21, 11:23 AM

you should sell that info .. gl

**RudyRuetigger** · 09-23-21, 11:28 AM

Originally posted by sweethook

you should sell that info .. gl

because he posted some plays that look like its from a computer?

what a dumb comment

im treating this guy like a tout until he proves otherwise

you are treating like someone knowledgeable until he proves otherwise

thats where we differ

**usma1992** · 09-23-21, 03:45 PM

I'm a Math/Comp Sci guy not a tout for sure

First, I may not get home in time to post NFL tonight in time. NCAAF system says Marshall ML... I think 40-29... NFL has CAROLINA 30-12 take Car... O/U is 43 so too close to call.

Now addressing my projected numbers versus bets. The projected numbers are calculations based on my algorithms. However, I have back tested these projections and sometimes certain aspects of either team forced me to eliminate the pick. So trust the pick more than the numbers but often those numbers come in really close.

I will post link to my app tonight. I would screenshot what it looks like but this sight won't allow pictures I don't think.

GL... I am anxious to see how it delivers this week.

Both... when I post NFL... you will see how close my numbers are. I'm very happy so far.

GL... Long day ... my high school went into lock down. Long story.

Dave

**stevenash** · 09-23-21, 03:48 PM

Interesting indeed.
Curious as to the end results.
Thanks for posting.

**usma1992** · 09-23-21, 07:23 PM

NFL Numbers

Time	AWAY TEAM	PROJ AWAY SCORE	HOME TEAM	PROJ HOME SCORE	TOTAL	VEGAS SPREAD	VEGAS O/U	SPREAD BET	O/U BET
8:20 PM	CAR	30	HOU	12	42	7.5	43	Away Minus	Too Close
1:00 PM	ARI	37	JAC	15	52	7.5	52	Too Close	Too Close
1:00 PM	IND	26	TEN	29	55	-5	48	Too Close	Over
1:00 PM	BAL	36	DET	29	65	7.5	49.5	Too Close	Over
1:00 PM	WAS	10	BUF	31	41	-8.5	45.5	Too Close	Under
1:00 PM	NEW	20	NEW	14	34	-3	41.5	Too Close	Under
1:00 PM	LAC	33	KAN	25	58	-6.5	56	Too Close	Too Close
1:00 PM	ATL	24	NYG	25	49	-3	48	Away Plus	Too Close
1:00 PM	CHI	16	CLE	21	37	-7	46.5	Too Close	Under
1:00 PM	CIN	15	PIT	18	33	-3.5	43.5	Too Close	Under
4:05 PM	MIA	13	LAS	32	45	-3.5	45.5	Too Close	Too Close
4:05 PM	NYJ	5	DEN	27	32	-10.5	41	Too Close	Under
4:25 PM	TAM	28	LAR	21	49	1	55.5	Too Close	Under
4:25 PM	SEA	26	MIN	31	57	2	55	Too Close	Too Close
8:20 PM	GRE	15	SAN	23	38	-3.5	50	Too Close	Under
8:15 PM	PHI	22	DAL	26	48	-4	52	Too Close	Under

Good Luck

**Tanko** · 09-23-21, 08:05 PM

Originally posted by usma1992

First, I may not get home in time to post NFL tonight in time. NCAAF system says Marshall ML... I think 40-29... NFL has CAROLINA 30-12 take Car... O/U is 43 so too close to call.

Now addressing my projected numbers versus bets. The projected numbers are calculations based on my algorithms. However, I have back tested these projections and sometimes certain aspects of either team forced me to eliminate the pick. So trust the pick more than the numbers but often those numbers come in really close.

I will post link to my app tonight. I would screenshot what it looks like but this sight won't allow pictures I don't think.

GL... I am anxious to see how it delivers this week.

Both... when I post NFL... you will see how close my numbers are. I'm very happy so far.

GL... Long day ... my high school went into lock down. Long story.

Dave

Thanks for the response.

Can you give a generalized description of what your algorithm uses as a basis for generating scores and an example of what aspects you may use to eliminate a game from consideration?

Also, you've mentioned the back-testing. What were the results when back-tested against data that was not used in generation of the algorithm?

I'm not looking for specific details, just hoping to gain some understanding and better feeling for gaining confidence using your system.
I realize it it likely complex w/ personalized tweaks so you don't want to provide information on it, I understand.

Look forward to your reply.
Good luck. I hope you hit 100% of your picks.

**Orbison** · 09-24-21, 12:01 AM

hi Dave, can you please post a link to your app when you get the chance? thanks for your replies and good luck this weekend.

**usma1992** · 09-24-21, 04:21 AM

Last nights... games. I had Marshall ML 40-29 versus APPI. Score ended 30-31. In my defense, Marshall settled for 2 field goals with 1st and goal from inside the 10 and missed a field goal to put them ahead. Appi's score was within two points.

Carolina versus Houston which I believe was the easiest NFL game of the week my system had Carolina 30-12. Carolina had 4th and 1 at the 4 yard line or so and didn't post any points. McCaffrey went down and I believe if he is in the game that is worth at least 6 points and a first down on the 4th and 1. Not a bad start.

Here is link: Only for Droids

Page Not Found

https://drive.google.com/file/d/1iwykZDMAePQ7njilG7p7KljQU2jKvZRn/view?usp=sharing

Web word processing, presentations and spreadsheets

When you download an APK file, it will ask you if you are sure. This has no virus but you might have to click yes a couple of times.
Regarding the algorithm. I primarily establish standard deviations, z scores and than calculate a projected total. However, certain combinations of zscores work more than others. Therefore, if I find that zscore combination to be failng despite what I feel are good predictions, the system tosses out that bet.

It is averaging 57%-60%. This week will be a real test. I have posted all my predictions and projected scores and bets so people can't say that I am making it up. It either works or it doesn't and I am not concerned about a specific week. If I hit 45% one week, I am ok with that. Below that number I will be concerned and revisit.
Dave

**usma1992** · 09-24-21, 09:00 AM

Let me know what you think of app... if u download.

Dave

**MrMiami** · 09-24-21, 10:19 AM

first, thanks for sharing
Question,
Why does your model say “too close” in the Fresno/UNLV game when your model predicts Fresno covering the -30 by 4+ TDs (59-0). Is there some other filter that is saying “too close”?

**usma1992** · 09-24-21, 10:34 AM

Yes there are other filters based on back tested data. I would consider Fresno personally but I'm only betting the games computer says to bet.

**username456** · 09-24-21, 11:02 AM

Good luck and thanks for posting!

**MrMiami** · 09-24-21, 11:04 AM

Originally posted by usma1992

Yes there are other filters based on back tested data. I would consider Fresno personally but I'm only betting the games computer says to bet.

Thanks for the response. I’ll be observing to see how your model does.

**LongBall52** · 09-24-21, 11:33 AM

Computer programs are a lot of help if one can narrow down the small window of play (s). They may not even be ANY in one week. If it had a high rate of success using ALL games, GOODBYE gambling on sports and horses.
Still a nice tool.

**usma1992** · 09-24-21, 10:23 PM

Looks like 2-2 tonight. Not great... not terrible. Expected Liberty to show. They had their chance. I will take a push for the evening. Tons of bets tomorrow. Too many to count.

GL. A Few of you have downloaded my app. Thank you for the feedback. It has been very positive. Now, let's hope I can hit 57% for the weekend.

Dave

**ttwarrior1** · 09-24-21, 10:48 PM

fresno is frauds though

**JohnGalt2341** · 09-24-21, 11:15 PM

Originally posted by usma1992

Honestly, trying to fine tune things for a year old two but it has been tested on back data. All I ask is feedback on how I can make it better. JJ I agree week 4 in college is early because of early matches. FBS vs. FCS... etc.

Two things happened this weekend that gave me hope. It picked NE Jets dead on 25-6... which is a very random score. And it picked Fresno St. ML to beat UCLA.

I will post NFL tonight and give you link, if u want to look at it. Only for droids right now.

I'm a Comp Science and Math teacher in high school so I got to get ready for class lol. But my degree is EE from West Point. All my fellow teachers are now using my picks. Apple users I just send spreadsheet.

GL... It has Marshall ML tonight and Carolina for NFL.

I'm only trying to hit 57% or greater.

Dave

Originally posted by usma1992

Last nights... games. I had Marshall ML 40-29 versus APPI. Score ended 30-31. In my defense, Marshall settled for 2 field goals with 1st and goal from inside the 10 and missed a field goal to put them ahead. Appi's score was within two points.

Carolina versus Houston which I believe was the easiest NFL game of the week my system had Carolina 30-12. Carolina had 4th and 1 at the 4 yard line or so and didn't post any points. McCaffrey went down and I believe if he is in the game that is worth at least 6 points and a first down on the 4th and 1. Not a bad start.

Here is link: Only for Droids

Page Not Found

https://drive.google.com/file/d/1iwykZDMAePQ7njilG7p7KljQU2jKvZRn/view?usp=sharing

Web word processing, presentations and spreadsheets

When you download an APK file, it will ask you if you are sure. This has no virus but you might have to click yes a couple of times.
Regarding the algorithm. I primarily establish standard deviations, z scores and than calculate a projected total. However, certain combinations of zscores work more than others. Therefore, if I find that zscore combination to be failng despite what I feel are good predictions, the system tosses out that bet.

It is averaging 57%-60%. This week will be a real test. I have posted all my predictions and projected scores and bets so people can't say that I am making it up. It either works or it doesn't and I am not concerned about a specific week. If I hit 45% one week, I am ok with that. Below that number I will be concerned and revisit.
Dave

I'm nearly certain this would make you one of the greatest cappers on the planet of all time.

What is your sample size so far?

**juicername** · 09-24-21, 11:18 PM

Originally posted by usma1992

Honestly, trying to fine tune things for a year old two but it has been tested on back data. All I ask is feedback on how I can make it better. JJ I agree week 4 in college is early because of early matches. FBS vs. FCS... etc.

Two things happened this weekend that gave me hope. It picked NE Jets dead on 25-6... which is a very random score. And it picked Fresno St. ML to beat UCLA.

I will post NFL tonight and give you link, if u want to look at it. Only for droids right now.

I'm a Comp Science and Math teacher in high school so I got to get ready for class lol. But my degree is EE from West Point. All my fellow teachers are now using my picks. Apple users I just send spreadsheet.

GL... It has Marshall ML tonight and Carolina for NFL.

I'm only trying to hit 57% or greater.

Dave

You should try to hit 100% and be happy if you hit 57%.

**usma1992** · 09-25-21, 03:17 AM

Appreciate the feedback...

I have data from 2012 forward excluding 2019 and 2020 because of Covid and I was about to break my computer. However, the step back away from the data proved to be beneficial. It allowed me to look at the numbers with a new perspective. I am a one man show.

So, here is what I do every week. I download all the Vegas spreads from a specific website into excel. It then reaches out to specific sites to grab all the data based on the current week. The toughest part is making sure the names match and the data is in the right place. Why? Vegas has Florida State like this and stats site has it FSU and results site has it Florida St. Vegas sites have it +24.5 which I have to turn into 24.5. The conversions are brutal especially for college basketball. However, that is all complete. I just have to adjust yearly.

After categorizing it, I have my statistical piece that is calculated then it produces my answers or desired results. I of course have to do quality control to look for bugs. For instance, my system couldn't recognize Ohio St. -46 because -46 had never occurred before in the data I sampled. Simple adjustment but nevertheless.

Finally, I save it to a separate worksheet and as a Comma delimited file. This then allows me to save that file to my Google drive where I have my app and the link to this weeks games. My app is immediately updated with the new data and the new games. After games are complete, I have to grab results from a different website that has different nomenclature.

The beauty of this is that it takes me 8-10 minutes or less. And if you doing it daily for baseball or college basketball or NBA, time does matter.

Now again, none of this means anything if it can't pick games correctly. However, I will say this. This year the people that are using my app are not always using my app to pick games, they are using my app to stay away from games.

Last week, it picked Fresno ML... this week it said stay away from Fresno. Personally, I would have bet it. I have posted my games and the link to my app if you can find a friend that has a droid. I am not asking for money just feedback. If you have a game that you think is way off, tell me and I will look at it.

As far as picking 100%, I have been doing this a while. I believe if you have a 100% as your upside, you have 100% as your downside. So far, my data has been fair. I am not into lieing ... doesn't make sense to me. I would rather be wrong and fix it.

So my philosophy is get my numbers as close to Vegas as possible then look for weaknesses. I don't want 100%. The real advantage I believe we have is that they have to produce a number for the masses on every game weekly. We can pick and choose which games we like.

Look back at my NFL games O/U are tight on Vegas lines except for a few.

Sincerely Dave

**usma1992** · 09-25-21, 03:32 AM

Reposting NFL picks... and highlighting system weaknesses...

System weaknesses:

#1 Injuries. There is no way for my system to account for Christian McCaffery going down. So the data will be based on what is happened to that point. If a major skilled position player gets hurt I would toss the next weeks pick.

#2 Weather. The two killers in weather in my opinion are wind and rain. If your betting the over, before you place that bet make sure it isn't 25mph winds especially for passing teams. If it a torrential downpour, than stay away also. I would also toss the heavily favored games like OSU-45. The rain effects OSU just like any other team. The might not cover because they won't score enough. Ironically, snow doesn't bother me. The defense slips and slides in snow so as players cut ... the defense has not ability to respond.

#3 NCAAF. Graduation of key players like Trevor Lawrence. My systems data pure to this year as of week 4. However, week 1,2, and 3 have a weighted piece to last years team. I think week 2 and week 3 are the most dangerous.

Anyway, good luck this week. Glad I didn't take Fresno. Liberty disappointed me.

NFL picks

Time	AWAY TEAM	PROJ AWAY SCORE	HOME TEAM	PROJ HOME SCORE	PROJECTED TOTAL	VEGAS SPREAD	VEGAS O/U	SPREAD BET	O/U BET
8:20 PM	CAR	30	HOU	12	42	7.5	43	Away Minus	Too Close
1:00 PM	ARI	37	JAC	15	52	7.5	52	Too Close	Too Close
1:00 PM	IND	26	TEN	29	55	-5	48	Too Close	Over
1:00 PM	BAL	36	DET	29	65	7.5	49.5	Too Close	Over
1:00 PM	WAS	10	BUF	31	41	-8.5	45.5	Too Close	Under
1:00 PM	NEW	20	NEW	14	34	-3	41.5	Too Close	Under
1:00 PM	LAC	33	KAN	25	58	-6.5	56	Too Close	Too Close
1:00 PM	ATL	24	NYG	25	49	-3	48	Away Plus	Too Close
1:00 PM	CHI	16	CLE	21	37	-7	46.5	Too Close	Under
1:00 PM	CIN	15	PIT	18	33	-3.5	43.5	Too Close	Under
4:05 PM	MIA	13	LAS	32	45	-3.5	45.5	Too Close	Too Close
4:05 PM	NYJ	5	DEN	27	32	-10.5	41	Too Close	Under
4:25 PM	TAM	28	LAR	21	49	1	55.5	Too Close	Under
4:25 PM	SEA	26	MIN	31	57	2	55	Too Close	Too Close
8:20 PM	GRE	15	SAN	23	38	-3.5	50	Too Close	Under
8:15 PM	PHI	22	DAL	26	48	-4	52	Too Close	Under

I love how close some totals are dead on and that some totals are very different. Phi/Dallas spread exact. Pitt spread dialed. Balt dead on. Cleveland dead on. Others not.

**usma1992** · 09-25-21, 03:49 AM

I have data from 2012 for most sports. Unfortunately, in NCAAB they changed the shot clock in 2015 so I have to treat that previous data worthless.

So for college football about 50*13 650 a year *10= 6500. I will continue to work. Next summer I plan to fine tune further. I believe I will never be done.

Hey, I just having fun and I know how difficult 57% is to achieve. But I am trying my butt off and I have the mathematical and computer knowledge to compete. At least I think so, lol.

**Tanko** · 09-27-21, 12:51 PM

I'm seeing the following results so far:

CFB
ML/Spreads 10-14
O/U 11-14-2

NFL
Spread 2-0
O/U 3-6
(Dal/Phi yet to be played):

I might be off a game or two since I pulled it together quickly. Looks like a tough week.
Let us know your stats.

Obviously one week does not make or break a system.