DraftAnalyzer7

[garion333]

Originally Posted by gstelmack

Okay, one more version pushed. If you jam all the sliders left, you should get a 0 score for the bars or combines as appropriate.

Nice, thanks. I've always kept them one tick over because with it all the way to the left there'd be some strange results (like -987.53 position scores and such).

[crusher78]

What do the white vertical lines represent in the bars section of draft analyzer? Are they related to combine ratings?

[corbes]

Can I put in a vote for someday changing the chemistry system from "predict affinity and conflict with position leader" to, more simply, "identify which of the four chemistry groups this dude is in?" I would find the latter much more useful as a screening device.

[gstelmack]

Originally Posted by crusher78

What do the white vertical lines represent in the bars section of draft analyzer? Are they related to combine ratings?

They are an attempt to do a rough matching of combines to bars. They are not particularly accurate, but they attempt to show you which bars you can expect to be high or low based strictly on the combine numbers.

[gstelmack]

Originally Posted by corbes

Can I put in a vote for someday changing the chemistry system from "predict affinity and conflict with position leader" to, more simply, "identify which of the four chemistry groups this dude is in?" I would find the latter much more useful as a screening device.

I could easily add something like that, but that's not something I can apply to a number to add in with the ratings.

The problem here though is everyone does chemistry differently, so there'd need to be some consensus on what exactly DraftAnalyzer should do.

[corbes]

No need for it to factor into ratings at all. All I want is a color code to know which of the four chemistry groups a player is in. I don't even care whether it knows what the conflicts are--just which of the 4 chemistry groups.

Ben's new page captures it perfectly--a colored background for the birthday is plenty, e.g., http://www.younglifenorthdekalb.com/...draftclass.php

Cuervo used to do something similar on the IHOF site.

[corbes]

Group 1
Aquarius -1 1-21 to 2-18
Libra -2 9-24 to 10-23
Capricorn -5 12-23 to 1-20

Group 2
Pisces -6 2-19 to 3-20
Taurus 2 4-21 to 5-21
Cancer 4 6-22 to 7-23

Group 3
Aries 1 3-21 to 4-20
Gemini 3 5-22 to 6-21
Scorpio 6 10-24 to 11-22

Group 4
Leo 5 7-24 to 8-23
Virgo -3 8-24 to 9-23
Sagittarius -4 11-23 to 12-22

[Nemesis]

Is there a workaround that will allow me to import existing player (not draftees) from like a Free Agent list, so I can apply weights to particular bars, so I can sort by a Bar Score?

[gstelmack]

No.

[gstelmack]

New version pushed with two notable features:

1) Added support for combine thresholds (see the discussion in What are the important combines for each position group? - Front Office Football Central. There is a new weight on the weights dialog that you can set to apply a penalty (default is -100) for any combine that is below one of the thresholds for the player's position. Set this to 0 if you don't want it. Combines below the threshold are colored red with black text (and fixed an issue where average combines could be colored red with black text when the player is drafted).

2) Added an option for "Color Chemistry Groups" as requested above. Uses Ben's web colors for the groups. Defaults to "on" so people know about it, but you can turn it off under the view menu.

[Nemesis]

In the old Draft Analyzer, there was a CombineData.csv - is there something similar with the new DA, or is it done internally?

[corbes]

Both new features are going to save me hours of time in preparing for drafts. MUCH appreciated Greg, thank you very much

[gstelmack]

Originally Posted by Nemesis

In the old Draft Analyzer, there was a CombineData.csv - is there something similar with the new DA, or is it done internally?

Still there, I updated it for the threshold values.

[Ben E Lou]

Booyah!

[Ben E Lou]

Originally Posted by gstelmack

Still there, I updated it for the threshold values.

Does it deploy to our local machines or does it stay on your server? (I'm not finding it.)

[Ben E Lou]

Well, either way, here's an update for FOF7.1a:

pos	AvgDash	StdDevDash	AvgSole	StDevSole	AvgBench	StdDevBench	AvgAgil	StdDevAgil	AvgBrdj	StdDevBrdj	AvgPosDrl	StdDevPosDrl
QB	4.791	0.15	27.86	7.1	9.5	2.2	7.869	0.25	102.6	6.0	71.2	9.5
RB	4.668	0.09	20.67	6.4	14.1	3.4	7.350	0.15	115.3	5.1	17.5	6.2
FB	4.812	0.08	21.87	5.7	19.7	3.9	7.584	0.20	103.9	5.1	23.5	7.5
TE	4.845	0.10	23.31	6.8	20.5	4.2	7.954	0.36	103.0	6.7	27.9	10.6
WR	4.559	0.10	21.04	6.9	9.8	3.1	7.263	0.19	108.0	4.9	40.4	9.8
C	5.350	0.13	29.55	9.9	22.6	4.8	8.119	0.22	88.4	6.7	0.0	0.0
G	5.293	0.13	28.12	9.7	26.7	4.2	8.019	0.27	90.0	6.3	0.0	0.0
T	5.320	0.16	27.23	10.3	26.8	5.0	7.915	0.27	92.2	6.6	0.0	0.0
P	5.087	0.18	25.68	7.4	10.1	3.6	7.692	0.34	106.9	7.0	0.0	0.0
K	5.153	0.14	26.98	7.4	9.2	2.9	7.672	0.33	103.8	6.4	0.0	0.0
DE	4.904	0.14	23.72	6.8	25.8	3.9	7.683	0.30	107.7	7.3	0.0	0.0
DT	5.139	0.14	22.35	6.4	27.4	3.7	7.919	0.33	97.3	9.5	0.0	0.0
ILB	4.884	0.11	27.01	7.1	20.2	3.5	7.677	0.24	104.3	5.9	23.3	7.7
OLB	4.748	0.12	25.16	7.1	16.8	3.5	7.458	0.20	108.6	5.8	23.5	7.6
CB	4.550	0.08	22.18	6.8	10.9	2.6	7.260	0.20	107.2	4.4	35.0	9.4
S	4.619	0.09	28.79	7.5	14.2	3.0	7.417	0.22	99.0	4.3	35.1	9.4

[gstelmack]

CombineData is installed along with the utility. It's tricky to find, though, as Windows installs ClickOnce apps in the application data area, not under Program Files.

Note that I changed it a long time ago to not use real averages and standard deviations, but instead to have average be the midpoint of the "black" range of combines, and "stddev" the amount needed to get to blue/green. That way DraftAnalyzer starts coloring combines when FOF does.

I'm open to comment on whether or not I should switch back.

FWIW, current version installed with Draft Analyzer attached.

[Ben E Lou]

Originally Posted by gstelmack

Note that I changed it a long time ago to not use real averages and standard deviations, but instead to have average be the midpoint of the "black" range of combines, and "stddev" the amount needed to get to blue/green. That way DraftAnalyzer starts coloring combines when FOF does.

Ahhhhhh! I didn't realize that. Then ignore the comments above for now. They were based on the combinedata.csv. I'll check the true averages and StDev from the CCFL's FOF2K7 era (30ish seasons) to get a better comparison.

[Ben E Lou]

Heh. All comments deleted. The combine averages (with correlation=50) are virtually identical from FOF2K7 and FOF7. I don't think it's worth spending more time to try to tease them out. Overall, clearly drafting QBs has changed significantly, as well as some other positions. However, from comparing the two datasets, its clear that the combines averages are unchanged between the two games.

[ahollz]

Dumb question, would there be any simple reason for the analyzer to not open at all? When i click on it a box opens stating "Verifying application requirements. This may take a few minutes." nothing develops after this and i was wondering if I could've done anything wrong. I was getting an error message stating player combine values could not be loaded so i uninstalled then reloaded everything and now i can't seem to get it to work.

[gstelmack]

Does that "Verifying application requirements" dialog ever go away? If not, try again in a bit, could be trouble with the website.

The combine values is a data file installed with the app - if you've ever customized it, it might not have been able to update it, but that would be unusual.

[ftwco]

I am very new to this program and so far like it alot, however could someone explain a few things:
Do I need to set weights or just use the defaults?
What is the columns titled Comb, rate and draft mean?
I assume that the green and positive numbers are the players you want to draft but I am seeing players that are way down on the grade showing positive numbers that have no hope of being drafted.
Please explain...

[gstelmack]

First post updated with some info for new players. Got a PM asking for the info now there.

[CentreCourt]

Hi,

Would like to try these excellent looking utilities.

Unfortunately having no success with either this DraftAnalyser nor the PlayerTracker. Both Utilities loaded on my system and run; just can't load any data !!!

In PlayerTracker I've loaded something as it says I can't do it again. I suspect I've not exported the required files from FOF7.

Draft Analyser currently giving me a big error message when trying to load draftees. The key one maybe that it cants find draft_personal.csv

I noted the following in the instructions

First, you must use the FOF options to Export Data and Export Personal Scouting Data during the FA period when the draft class is available. This is the data the utility will use.

I can only see Export Data as the last entry on the Almanac Menu

Any suggestions much appreciated.

[Ben E Lou]

Have you updated to the latest version? The draft_personal.csv was not available in the release version. It showed up in a patch.

[CentreCourt]

Originally Posted by Ben E Lou

Have you updated to the latest version? The draft_personal.csv was not available in the release version. It showed up in a patch.

Hi Ben.

That was exactly it. I didn't realise I'd never installed a patch (not that I ever checked my version). PlayerTracker now working an I'm sure DraftAnalyser will once I reach a draft.

Many thanks.

[CentreCourt]

Having great fun with this Utility. Really impressive.

I wondered if somebody could help me understand how the 'scoring' works on Combines.

I've set the Combine Weight to 10 and for every position I've set Solecismic to Max and all other Combine activities to Min.

For Tackles, a Solecismic of 22 scores zero. Each Solecismic step of 1 scores 2. So 47 scores 50, whilst a 10 scores -24.

For QB a 27 scores zero. A 28 scores 2.31 and then it increments in 1.54. So a 33 scores 10. Interestingly 26 and less all score zero, there are no negatives.

Now I understand a QB 28 is the sweet spot for QB's. Below and 50/50 is the best you are likely to get. Equally Solecismic appears to have no influence on Tackles true ability.

The reason for this 'strange' experiment was I was trying to understand first principles in order to set my own Weights. I just chose Solecismic first as the 1 through 50 scores appeared easiest to correlate.

But now I'm confused. I thought the influence of Solecismic was based entirely on where I positioned the dial. But it appears there are also hard coded values affecting this. Can I see these values anywhere? And also can anybody explain why there is an inbuilt Solecismic value per position?

Any thoughts appreciated.

[aston217]

That's interesting, I didn't know this was how DraftAnalyzer worked. That's cool and probably how it should work.

The reason for that probably comes from MalcPow's thread of combine benchmarks for FOF2k7 (Some Combine Benchmarks (and Drafting Thoughts) - Front Office Football Central. No player who was set to have 60 Avoid Intelligence (as well as 60 in the other set-able bars across the board) scored below a 28 in Sole. It's less of a 'sweet spot' as it is a 'cutoff.' More knowledgeable people than me may be able to tell you concretely whether we should consider these numbers as still valid in FOF7. So far, I'm rather ASSuming it.

However, as far as cliffs go, that isn't the worst. Most other "cliff" - like scores are much more stringent. For example, if an OT scores worse than 7.80 agility, MalcPow's benchmarks strongly suggested he has no chance of greater than 20 in pass blocking. If DraftAnalyzer accounts for this, that's great, because a 7.82 and a 7.80 both appear "black" in game -- nothing to suggest such a disparity in quality.

Off the top of my head, I can't think of any 27 or lower sole score QBs who really panned out. There was one in the CyFL who reached modestly high OVR -- in the 50s -- but he seemed to be a pick machine, despite his very low intelligence. That may have been the offense or receivers too, who knows.

[irabowman]

Thought I'd try this tool out did everything was suppose to do. I get a IO Error: about the file not being found. Other day I learned I have something called hidden files. Will I not be able to use this because of the hidden files?

[garion333]

Originally Posted by irabowman

Thought I'd try this tool out did everything was suppose to do. I get a IO Error: about the file not being found. Other day I learned I have something called hidden files. Will I not be able to use this because of the hidden files?

When doing what?

[gstelmack]

Basics: You need to export both sets of data from FOF: Export Data and Export Personal Scouting Data from the Almanac window. Do it when FA starts (when you can submit bids for FA1) and not before as the draft class won't be available.

As for combine scores, the utility takes the standard deviation from average and multiplies it by the combine weight. The combine weight is a mix of the global weight for combines and the specific weight of that particular combine you gave in the sliders on the right. Basically all the bar values are added up, each individual bar is divided by that total (so total weight = 1.0), that is multiplied by the global combine weight value, and then your standard deviations (adjusted by a factor) are multiplied by that. The "adjusted by a factor" basically means larger standard deviations are more important so I actually take the standard deviation and raise it to a power for a curve instead of a line. So, a stddev of 2 might actually be 2.8 for the math.

Then you have the threshold modifier that lets you set a penalty for anyone below a "cliff" threshold. That's just a straight "subtract X for every combine below the cliff threshold for that position".

[Nemesis]

Originally Posted by aston217

That's interesting, I didn't know this was how DraftAnalyzer worked. That's cool and probably how it should work.

The reason for that probably comes from MalcPow's thread of combine benchmarks for FOF2k7 (Some Combine Benchmarks (and Drafting Thoughts) - Front Office Football Central. No player who was set to have 60 Avoid Intelligence (as well as 60 in the other set-able bars across the board) scored below a 28 in Sole. It's less of a 'sweet spot' as it is a 'cutoff.' More knowledgeable people than me may be able to tell you concretely whether we should consider these numbers as still valid in FOF7. So far, I'm rather ASSuming it.

However, as far as cliffs go, that isn't the worst. Most other "cliff" - like scores are much more stringent. For example, if an OT scores worse than 7.80 agility, MalcPow's benchmarks strongly suggested he has no chance of greater than 20 in pass blocking. If DraftAnalyzer accounts for this, that's great, because a 7.82 and a 7.80 both appear "black" in game -- nothing to suggest such a disparity in quality.

Off the top of my head, I can't think of any 27 or lower sole score QBs who really panned out. There was one in the CyFL who reached modestly high OVR -- in the 50s -- but he seemed to be a pick machine, despite his very low intelligence. That may have been the offense or receivers too, who knows.

[aston217]

Ha! That should read "Avoid Interceptions."

Avoid Intelligence is the bar I got the most of

[da0ud]

Hi guys,

Does anyone of you know in which of the csv files I can find the individual bars of the players which are in the amateur draft ?
I found their combine values in the rookies.csv file but the individual bars are missing.

Thanks a lot,
Daoud

[gstelmack]

draft_personal.csv. It's only generated when you tell the game to export personal scouting data, and only during FA1 - the draft class is generated at the first stage of free agency, and goes away once the draft starts. Basically, "View Draft Class" must be available in the large buttons in the upper left of the UI for this file to be generated by the game.

[da0ud]

Originally Posted by gstelmack

draft_personal.csv. It's only generated when you tell the game to export personal scouting data, and only during FA1 - the draft class is generated at the first stage of free agency, and goes away once the draft starts. Basically, "View Draft Class" must be available in the large buttons in the upper left of the UI for this file to be generated by the game.

Thanks a lot gstelmack, I found them

I am looking at making some cool tools as well.
I hope one day I can have complementary tool to yours and hopefully as nice looking as yours.
For the moment I am only playing around and see what info I can gather.

Cheers, Daoud

[gstelmack]

Added this info to the first post, but pulling it here for those viewing unread:

UPDATE March 16, 2015, v 1.2.0.0:

Today's update has a new way of viewing the draft class based on some data mining analysis. You will see 5 new columns in the table view:

D? - how likely is this player to develop to a 40 rating or higher?
VG - how likely is this player to develop to a Very Good or Excellent rating?
G - how likely is this player to develop to a Good rating?
A - how likely is this player to develop to an Average rating?
FP - how likely is this player to develop to a Fair or Poor rating?

The calculation for these new columns is a combination of the player's grade, combine scores, percent developed, and the min range of the blue bars. The latter is scout dependent, so these ratings will not be the same for all scouts. Remember that these are percentages - my test cases showed in general about a 50% accuracy on the highest chance range being the range the player fit into, and above 95% on being within 1, if you've got a good scout. You will have to look at the pattern, and beware of players that are like 20/30/30/20 chances on these (QBs especially love to be unpredictable).

In addition, there is a new command on the Edit menu called "Slot by Rounds". This works by calculating a "slot score" from these new values (it's shown in the player details window). It's relatively straightforward, using the middle of each of these ranges to estimate player development, with D? being given a 40 weight (VG is 85, FP is 10, etc). The slot score is multiplied by your assigned position weights, and then banded to give players 1st, 2nd, 3rd, etc round slotting. That generates a draft order for all players. There is a special case for high return and special teams bars that can percolate otherwise poor players into the 7th or possibly 6th rounds.

This should help bring some improved advice to the drafting, building in some of the patterns expert drafters have dug up on their own, but remember that drafting is still playing the odds, this may just help you understand the odds a bit better.

[gstelmack]

One other update in here - "Export Draft Order" will now skip drafted players, so no need to update the draft order after marking draftees to get your top 100.

[FootballDownUnder]

Hey Greg, great work! Going into my first ever MP draft and while I'm still getting my head around it, your tool is already invaluable.

[aston217]

What new devilry is this!

Greg, longshot here, but is there any chance you would be willing to share the functions that calculate those values? I still like taking notes the most on each draft prospect, so I don't know if I will use DA regularly, but I'd probably incorporate it into my version of it if I could. That's some pretty advanced stuff.

[gstelmack]

I'm using the Accord.NET framework for the hard math bits, the tricky part is documenting the input data (converting bars / combines to the correct Z values) and spitting out the coefficients. I should probably push a source code release at some point.

[corbes]

Why doesn't slot by round eliminate all of the below-threshold combine players? Are the findings that those players aren't quite so bad as originally thought?

[gstelmack]

A few points on thresholds:

1) The threshold is determined by how likely they are to get better than 50. You are GOING to be drafting sub-50 players at some point during the draft, there just aren't 224 50+ players in a draft.

2) There are exceptions on the thresholds, just not that many.

3) I just let a learning algorithm run on the data, letting it figure out thresholds on its own.

I suspect the only below-threshold guys you'll find high in the draft list have other bars that make up for it

[Ben E Lou]

Originally Posted by corbes

Why doesn't slot by round eliminate all of the below-threshold combine players? Are the findings that those players aren't quite so bad as originally thought?

Different people look for different things. Having observed your drafting lately, you and I are similar in that we're looking for 50+ types all draft long, through round 7. Some people look for more "specialty" types as the draft goes longer. When it comes to what Greg mentioned in terms of "playing the odds," for me, it's just not worth the extra time and effort to do a bunch of digging for non-constraints players who are 50+. For example, you'll only find a 50+ Center who is below the dash constraint once every 30+ years, below the bench constraint once every 5 years, and below the agility constraint once every 7 years. Sure, you might find a 48/48 guy who fits the bar profile you like, and those guys are rightly showing up. But until the MP collective improves its drafting, in most leagues I don't think you need to bother with below-constraints guys.

That being said, Greg is in only one MP league--FOOL. And FOOL is--by FAR--the best-drafting league when it comes to snatching up constraints guys. (There are typically less than 5 guys I'm still interested in there post-draft. In other leagues, that number is frequently in the 15-20 range.) As a result, you NEED to draft non-constraints guys there, so there's a very good reason for him to include them.

[Ben E Lou]

Originally Posted by gstelmack

1) The threshold is determined by how likely they are to get better than 50. You are GOING to be drafting sub-50 players at some point during the draft, there just aren't 224 50+ players in a draft.

Yeah, see my post below yours. Your one MP league is the best collective set of drafters that exists in any league I'm in, and very likely in all of FOF MP. I have been able to get 50+ players in the 6th and 7th rounds and even as undrafted free agents elsewhere--just not in the FOOL. Not even close.

[gstelmack]

This is also why I didn't replace the old numbers, but just added to them. I fully expect people will filter and redo these, either not using slotting at all and instead using the new info to help them tweak things, or going back through after the slotting and manually adjusting the slotting of some players. Or ignoring it altogether.

[garion333]

Wow, I am getting some funky results with the round grading (slot by round). It appears no combines may be sneaking way up.

Even after resetting my weights this guy comes out on top:



Maybe I'm way off, but I don't buy that.

Help? Thoughts?

(This is from the 2037 draft in the OSFL.)

[gstelmack]

No Combines = use bars + multiply by position weight. That LDE has pretty good bars by your figures, and I suspect you have a high position weight for DE. I will take a look at the multiplier I use to get from a bar score to a slot weight, though, it's possible I'm not properly handling the relative value of using min vs avg vs max bars to determine the bar weight.

[garion333]

Originally Posted by gstelmack

No Combines = use bars + multiply by position weight. That LDE has pretty good bars by your figures, and I suspect you have a high position weight for DE. I will take a look at the multiplier I use to get from a bar score to a slot weight, though, it's possible I'm not properly handling the relative value of using min vs avg vs max bars to determine the bar weight.

I reset to defaults, so this is your weights. But, yes, overall I would have high DE ranked pretty high.

[garion333]

Shouldn't his light green Pass Blocking be Red since it's outside (7.85) the constraint/cut off (7.80)?