Estimating Weekly FAAB: Step One
A couple weekends ago, Dave Appelman asked my fellow writers and me what features we’d like to see available at RotoGraphs (more additions are in the works) and a FAAB (free agent acquisition budget) bid estimator was discussed. With so many leagues and so many bidding setups, I accepted the daunting challenge. I’m beginning to produce some limited usable results.
Before I go into the many inputs in FAAB bids such as declining budgets and absentee owners, I needed:
1. A way to measure owner seniment before the normal Sunday FAAB bids.
2. A reliable source of FAAB bids.
For the first source, I’d like to thank fellow RotoGraphs contributor, Al Melchior for pointing out that CBS Sportsline displays historic weekly ownership rates. Every host website will display ownership rates. What was important with this source is they have a ton of quick grab waiver wire leagues. As soon as a player bursts on the scene, owners can go pick them up off the wire. The ownership rates for these hot commodities immediately begin ticking up. These values are available but the bases need rounding a few times to collect them.
First, an owner needs to set up a CBS league. I can’t seem to find a way around this requirement. If anyone does, let me know. Next hover over the “PLAYERS” tab and select “STATS”. Then search by the player’s name and right click on it. This popup will appear.
From this window select the “FANTASY TRENDS” tab where the weekly ownership rates are shown.
With the help from CBS Sportsline’s Heath Cummings, I determined the ownership rates lock right before FAAB runs that midnight. It’s a perfect source of ownership rates even though they are little tough to acquire.
As for the FAAB bids, I’m going to owners who have thousands of dollars on the line and am using NFBC’s Main event bids. Thanks to NFBC’s Greg Ambrosius for letting us use the bids. The winning bids are from 34 leagues with 15 owners in each league with thousands paid in entry fees. These owners are trying to win.
The NFBC message board contains all of the weeks’ bids ($1000 budget) which can be matched up with the trends on CBS.
I decided to start my analysis with just week 14’s bids. I had no idea what useful information I was going to be able to find. And if I did find anything, I might as well make it useful during the current time frame.
I collected the average and median winning FAAB, the standard deviation in FAAB, the ownership change amount, and overall ownership rate. One item I found was that the largest bids were for single players dropped my other owners (e.g. Mark Trumbo). It was a tough call between using the single large values and those bid on the most. I went for a middle ground. I collected information on a few dozen players.
I quickly found the players needed to be divided up into two groups, possible closers (higher than average bids) and everyone else.
I tried several combinations and in the end, I found comparing FAAB bid to ‘ownership percentage’ plus ‘weekly percentage jump increase’ made a great starting point. I might need to weight each value in the future.
It is just a starting point and I may need to split the data into more groups (e.g. rookies, starters, etc) and refine the equations. What I was able to find were a couple of equations to estimate week-14 FAAB values.
Potential closers: $60.3*(Total ownership + ownership increase) + $8.4
R-squared: 0.53Everyone else: $9.1*EXP(2.12* (Total ownership + ownership increase))
R-squared: 0.52
In each instance, I compared the standard deviation to the actual bid to find the normal range of outcomes. In each instance, the range was +/- (0.6 * expected FAAB value).
For an example, this week Enrique Hernandez, who is playing better, saw his CBS ownership jumped from 7% to 33%. So, the constant used with the above formula would be 59% (33% + (33% – 7%)).
$9.1*EXP(2.12* (.59)) = $32
Additionally, the winning bid range was projected between $13 and $51. In nine bids, he went for a little less for an average $26 bid. All but one value was in the $13 to $48 range.
Overall, it’s an encouraging start. That’s it though. It’s not useful to AL or NL only. Yet. Or smaller or deeper leagues. What it has done is encourage me to take another step and begin to refine the formula some more.
I plan on just increasing the bid sample size over the next few days. In the meantime, please let me know if there is any additions or changes I could do. I’ll be back in a day or so with some more information.
Jeff, one of the authors of the fantasy baseball guide,The Process, writes for RotoGraphs, The Hardball Times, Rotowire, Baseball America, and BaseballHQ. He has been nominated for two SABR Analytics Research Award for Contemporary Analysis and won it in 2013 in tandem with Bill Petti. He has won four FSWA Awards including on for his Mining the News series. He's won Tout Wars three times, LABR twice, and got his first NFBC Main Event win in 2021. Follow him on Twitter @jeffwzimmerman.
Gooderly!
I might have missed it, but did you check your equations across various weeks yet? I would guess that generally declining FAAB budgets might have an effect on how FAs are assessed. It wouldn’t suprise me, even, to find out that certain positions might gain or lose value at different points in the season.
After playing for a million years in non-bid leagues, I’m in my first ever FAAB league this season. I find trying to guess how other owners are going to value pickups to be maddening…
Did you read the article? I tried to answer these questions.
*lol* I did. Basically I’m asking for unfounded speculation based on incomplete research that you may or may not have already begun 🙂
I just wanted to see how it looked this week and then go back further. Baby steps
If I understand correctly, it seems that you are collecting the data from CBS by doing individual searches on the Players->Stats page, clicking on the player name and then clicking on the Fantasy Trends tab on the resulting player page, followed by estimating the ownership values by eyeballing the resulting chart that appears. That seems too labor intensive.
I assume the CBS guys have made you aware of the Roster Trends page. If not, you need to check them out because they can provide much (all?) of the data you want in just a few tables.
On the Roster Trends page, you can sort on “Most Added/Dropped” to list the 50 most added and dropped players for the week, with the previous week ownership percentage, current week ownership percentage and difference all in one easily copy and pasteable table – you can also sort the data by position if that is helpful (but not RP vs SP unfortunately).
You can also use the Roster Trends page to sort by Most Owned, which lists the 100 most owned players with their current ownership percentage (as well as current starting percentage, which could be useful to determine which players are being picked up for streaming purposes).
To be clear, you can populate the Roster Trends tables using only free agents in the league you are in. so the 100 Most Owned table in the Roster Trends page shows you the 100 players that are most owned *in leagues other than your own league*. It is an excellent tool for identifying the players who an owner might want to pick up in FAAB.
That table is fine for this week but does no good for past weeks.
True – it looked to me like what you described doing was only using this week’s data. (And I would think any final algorithm would use only current week data – though certainly it would be advisable to validate for past weeks’ data as well.)
One other group for which you might need a separate formula is pitchers expected to have two starts the following week – that is a distinguishing FAAB bidding factor for many streaming pickups. CBS identifies those pitchers on their “2 Start Pitchers” page, but only for the scoring period following the current scoring period. I am not sure how you would identify those pitchers for past weeks (you could use actual two start weeks as a proxy, but that is subtly different than the expected starts that the FAAB bidders were using as their data).
I like what you’re trying to do here (using CBS trends is how I plan my bids each week) but I don’t see it being robust enough to handle the diversity of leagues. The current formula would always suggest a minimum bid of $9. I’d expect you’re much better off churning your roster with $1 pick-ups throughout the year. I think a different approach that may work well would be use the ROS projections with the auction calculator. Then work on finding a formula to transform these numbers to a weekly bid. Admittedly I don’t think this would work well with closers or prospects (which is probably the focus). Looking forward to see where you go with this!