Archive for the ‘Theory’ Category

« Older Entries

Does this Tool Exist?

25 Comments
March 16th, 2011 by
Categories: Strategy, Theory

Lately, I’ve been imagining a tool that would offer in-draft advising about which fantasy player to draft. I’m curious if anyone else thinks this sort of tool would be useful and if anyone knows of an existing fantasy tool that does this.

Essentially, I’m interested in two things during the fantasy draft:

Tell me how much each available player helps my team. I imagine this in terms of Standings Gain Points (SGP) — how many points in the standings would I gain if I added this player to my team? If it knows which players each team has already drafted, these would be SGPs customized for my particular team and situation. And so it presents a list of players ranked by their SGP benefit to my team.

At first, the players with high dollar values will also be the ones with the biggest SGP benefit. However, I could see things shifting as the draft progresses. Maybe I’m weak in stolen bases, so that drafting Juan Pierre could move me up 5 SGP. Pierre might not be the highest valued player on the board in terms of dollar value, but he could be the player with the biggest benefit to my team in this particular situation.

Pierre’s dollar value — calculated for an average team in an average league — could still be static. But I might be willing to go a few dollars over his book value if this tool tells me I really need SB. Or I let a similarly valued player like Carlos Lee slide (even at a low price) so that I can save an OF spot for Pierre. Or I see that Pierre, Brett Gardner, and Rajai Davis would all provide a good SGP boost to my team, so I don’t need to spend any extra as long as I get one of them.

Potentially, all of these data could be factored in to actually adjust Pierre’s dollar value. My initial, simplified idea would just present an assessment of the pros and cons and let the drafter decide how to use that information.

Tell me how much each available player helps each other team. Secondarily, I’m also concerned about keeping players from my rivals. What hurts my opponents may help me, even if there is no direct SGP benefit for my team. I might draft a player who gives me no SGP if that ensures he stays off the team of my closest opponent (where perhaps he would give 5-6 SGP and let that team pass me in the standings).

And so I imagine the tool listing out — alongside the SGPs each player contributes to my team — the SGPs that each player would give every other team.

Those two things seem like they would be pretty simple to implement. And so I’m curious:

Would this tool actually help in a fantasy draft or auction?
Does anyone know if there’s already a fantasy product out there that does this?
What other information might this type of tool factor in to its recommendations?

More Math from Mass

6 Comments
February 27th, 2010 by
Categories: Fantasy Basics, Other Sites, Theory

I wish I could resist the lure of AJ Mass’s ESPN.com articles, but I cannot.

Mass answers the question I’ve always been wondering: “What happens if you rank fantasy hitters using bizarre, illogical criteria?” His top 10 fantasy hitters:

1 Jacoby Ellsbury
2 Carl Crawford
3 Michael Bourn
4 Albert Pujols
5 Matt Kemp
6 Mark Reynolds
7 Ryan Braun
8 Ian Kinsler
9 Hanley Ramirez
10 Chone Figgins

Now there’s something unusual about that list, but I can’t quite put my finger on it… Mass has already expressed his love for Ellsbury, so I guess it’s not too surprising that he puts Carl Crawford #2 and Michael Bourn #3.

How did this happen? The secret is in the faulty starting assumption:

What we’ve done is very simple. We’ve taken the league-wide totals from last season to create a statistical universe for our players to inhabit. We determined the expected statistical output of the average player, assuming even distribution among the lineup spots. From this, we were able to extrapolate the relative value of each hit, each run scored and so on. In other words, since there were more home runs than stolen bases in 2009, by a ratio in the neighborhood of 7-to-4, the value of each stolen base was about seven-fourths that of a home run, matching the relative frequency of the event.

I’m pretty sure you want to use the totals from a typical fantasy league, not all of MLB. Using the pool of fantasy starters will either get you to SGP or standard scores, either of which should yield a pretty realistic result.

Using the pool of all MLB hitters is just going to give you a mess.

What is Chone Figgins Worth at 2B?

7 Comments
February 26th, 2010 by
Categories: Position Eligibility, Theory

The Seattle Mariners started off Spring Training this week with a surprising infield arrangement — Jose Lopez taking grounders at 3B and the newly acquired Chone Figgins playing 2B.

At this point it just looks like an experiment, but the potential shift has lots of people talking. For fantasy purposes, a 2B-eligible Figgins would be a throwback to his legendary flexibility of years past.

But how valuable is Figgins at 2B? As a 3B in a standard league, I’ve got him at $11. That may seem a bit conservative after his breakout $19 campaign in 2009, but don’t forget about a 2008 where he barely managed to be above replacement level. He’s also making the transition from the Angels to the Mariners, and I don’t see many guys in Seattle that will be able to provide much help for his run totals. He’s hard to predict for 2010, but the best guess is that his true talent lies somewhere in between 2008 and 2009 — maybe about $11.

All of that assumes he only qualifies at 3B for fantasy, though. You might notice that the Price Guide lists a “Total” value and an “Adj. Total.” The former represents a player’s value without regard for position; the latter is after taking position into account. Since the typical 2B is a little worse than the average 3B, a player with equivalent stats will be worth a little more as a 2B than as a 3B. The “League Info” tab of the Price Guide tells you how much each position is adjusted.

To figure out what Figgins is worth as a 2B, all you have to do is replace the 3B adjustment with the 2B one. It turns out that, in a standard fantasy league, Figgins would get a bump from about $11 to $15 if you could put him at 2B.

There’s also some value in a player qualifying at two positions: You gain roster flexibility for off days, and you have more options for replacing an injured player. Those things are tough to quantify, but I’d pay a couple of extra dollars for a 2B/3B. I could easily see a swing from $11 to $18 when everything is accounted for.

Keep in mind that all of this is for a standard league that starts an extra CI and MI. Surprisingly, in a shallow league like the Yahoo or ESPN default setup, there is basically no gain for switching from 3B to 2B. The hypothetical 2B-Figgins is valued at $8, just a slight improvement from the expected $6. With only 10 or 12 starting at each position, the drop-off on 2B is much closer to that of 3B than to SS.

In a situation without a MI slot, I’d be willing to give a little extra for multi-position eligibility. So maybe that original $6 becomes $12 if Figgins is a 2B/3B in Yahoo/ESPN.

AJ Argues for Ellsbury in the First Round

5 Comments
February 19th, 2010 by
Categories: Fantasy Basics, Other Sites, Theory

AJ Mass at ESPN explains how Jacoby Ellsbury should be a first round pick this year.

Regarding those who might prefer Prince Fielder or Ryan Howard in the late first round, AJ says:

They might have mainstream public opinion on their side, but you would have mathematics. Allow me to explain why you just can’t let him get past you with, say, the seventh, eighth or ninth pick in the first round and leave open the possibility that he might not make it back to you. You need to take him with that first-round pick.

Well, who am I to argue with mathematics?

AJ’s idea is what he calls “Mass Effect,” a valuation system that makes Ellsbury the #1 overall player in fantasy last year. That’s right: Jacoby Ellsbury, Fantasy MVP.

Did I mention that it ranks Carl Crawford #2, Derek Jeter #5, and Ichiro Suzuki #7? Something doesn’t seem right here…

Well, AJ acknowledges that Mass Effect has “one flaw.” He explains how stolen bases become less valuable throughout the season due to diminishing returns. So to adjust for the changing value of stolen bases, he decides to knock off 50% of the value of each stolen base.

I’m sure he went through some of his extensive mathematics to come up with that 50% figure, since it drops Ellsbury from #1 overall to a much more realistic #10 overall.

Just to recap:

1. Come up with a player valuation system that doesn’t make sense.
2. Instead of recognizing that the system doesn’t make sense, fudge the numbers some so that they look better and yet still fail to reflect reality.
3. Call it “mathematics.”

I love it.

How Valuable is Multi-Positional Eligibility?

1 Comment
February 10th, 2010 by
Categories: Fantasy Basics, Theory

I commented earlier on players that will be eligible at multiple positions in 2010. I think everyone would agree that there is fantasy value in being able to shift a player to a new position. The question is determining how much that ability is worth.

While I don’t have a precise formula for putting a dollar amount on extra eligibility, I think there are several important factors:

Some extra positions are more valuable than others.
Victor Martinez may qualify at both C and 1B, but his value at C is so much greater that it isn’t that useful to be able to put him at 1B.

As far as specific positions, I’d guess that the most valuable are players who qualify at both SS and 2B, like Asdrubal Cabrera. The next most valuable are middle infielders that also qualify at 3B, like Ian Stewart or Jhonny Peralta.

Less valuable are the players who just add 1B eligibility. Not only are there plenty of regular 1B available, but 1B is also the most common extra position in fantasy.

Multiple eligibility is less valuable in leagues with CI/MI positions.
Traditional fantasy leagues start both a “corner infielder” (either 1B or 3B) and a “middle infielder” (either 2B or SS) in addition to the actual baseball positions. These “flex” positions give you some natural wiggle-room: If your 1B gets hurt, you can maybe slide your CI to 1B, and replace your CI with either a 1B or a 3B.

Since you can use either a 1B or a 3B, in this situation a 1B/3B player is less important. But this makes a huge difference in a standard Yahoo or ESPN league that only starts one infielder at each position.

A multi-positional player is more important in a league that starts fewer players and less important in a traditional league with CI and MI.

Multiple eligibility is less valuable in leagues with transaction limits.
If your fantasy league lets you make daily roster changes and doesn’t cap games played, a multi-positional player is great. The simple ability to fit players in the lineup more often means you have an advantage in counting stat categories like HR, SB, R, and RBI.

However, if your league only allows weekly transactions, this strategy doesn’t work as well. Really, a multi-positional player just gives you a little extra coverage when you have a player go down with injuries.

All told, I’d probably be willing to spend an extra $3-4 for a multi-position player, or maybe draft them a round early. But that evaluation is very dependent on the context.

Adjusting the Hitter/Pitcher Split (Part III)

2 Comments
January 22nd, 2010 by
Categories: Strategy, Theory

Part I of this series was simply a look at a new feature of the Price Guide — the ability to customize the hitter/pitcher split whereever you wanted. Part II was an experiment to see how that feature works in real-life, using the Tout Wars league as an example. Now we get to the implications of this tool — Should you create values that imitate the values of your league?

Last year, I spent quite a bit of time trying to figure out why real-life drafters were willing to spend more on hitters and less on pitchers than what the Price Guide recommends:

Examining the 70/30 Split
Diverging from ADP
More Thoughts on ADP

My final conclusion was that…well, I never really figured out why this happens.

However, I’m not sure if a conclusion on what is most correct is really essential. I have two simple things in mind for approaching my drafts this year with regard to the hitter/pitcher split:

1. Adjust the prices to fit the league’s historical tendencies.
This would mean, for example, that if I were in Tout Wars Mixed, I would use the 72/28 split that models reality instead of the 63/37 that the Price Guide default spits out. I don’t know if this is the correct thing to do, but it does work.

Why is that? In simulating drafts, I’ve noticed that trying to buck league trends is a quick path to failure. If you are willing to draft pitchers early but the other eleven people in your league wait, you will probably lose.

The opposite is also true — if everyone in your league drafts pitchers with their first picks and you draft hitters instead, you will also lose. One person by himself or herself cannot exploit a league’s weakness. A player is only worth what the league is willing to pay for him.

Adjusting the split to match the league is valuable because it helps get the bargains. If I get in a bidding war for CC Sabathia and win him at $40, I might only be paying a fair price according to the Price Guide defaults. But if the bidding on Sabathia is an outlier and most top pitchers go for around $30, then I am probably missing a bargain price on someone else because I was willing to pay a “fair” price on Sabathia.

If I adjust the split, I will either pay a fair price (according to your league) or a bargain price (according to the default Price Guide split). Either way, it is impossible to overpay.

2. Be willing to spend extra on pitching when necessary.
I balance the first strategy with a willingness to occasionally exceed the adjusted split.

While I’m not really confident in the Price Guide’s default splits, I’m also not confident that the leagues are doing a better job. And so, for certain pitchers, I’m going to be willing to spend an extra $1-2.

Doing so will still leave me below the Price Guide’s default prices on those pitchers, and it still allows me to bid competitively on the top-tier hitters. Once again, I think it’s as close to a win-win as I can get.

Am I on the right track here?

Adjusting the Hitter/Pitcher Split (Part II)

1 Comment
January 21st, 2010 by
Categories: Theory

Yesterday, I described a feature that lets you force the Price Guide to use any hitter/pitcher split you desire. Even if the Price Guide suggests a 67/33 split, you can make it go 70/30 instead.

Let’s try out the enter your own hitter/pitcher split feature using the Tout Wars Mixed League. Tout Wars is a competitive expert league using standard 5×5 rules, with 17 teams in the Mixed league.

At the 2009 auction, the highest paid hitter was Jose Reyes at $55. (Paying $55 for Jose Reyes was not a recipe for success last year, by the way.) The most expensive pitcher was Johan Santana ($34). The names and exact dollar amounts aren’t really important, though; we’re just trying to identify the basic scale for hitters and pitchers. In general, hitter values top out at $50+ and pitchers at $30+.

The Price Guide disagrees radically — the 2009 composite projections with the Tout Wars settings ranked the top hitter at $45 and the top pitcher at $47. Instead of the $50+ and $30+ that occurred in reality, we’re looking at basically $45+ and $45+. The top pitcher has not only closed the gap with the top hitter, but actually surpassed him.

Looking across the top prices, the Price Guide’s top hitters are all too low and the top pitchers are all too high using the default settings. The problems stem from the hitter/pitcher split that the Price Guide settles on: 63/37. In real life, Tout Wars participants are willing to spend considerably more on hitters and are quite a bit more conservative with pitchers.

What can we do to align these perspectives? Let’s start adjusting the Price Guide’s hitter/pitcher split beyond the 63/37 it start with. By the time we get to 72/28, things are looking better: The top hitter is now $51 and the top pitcher is at $35, pretty close to the $55 and $34 we started out with. More importantly, the lists are congruent when compared past the top spot. The 10th hitters ($37, $37), 20th hitters ($30, $31), and 10th pitchers ($20, $20) are all quite close in value between the Price Guide and the actual results in Tout Wars.

So now it’s pretty easy to make sure that the Price Guide’s results conform to reality when you are generating dollar values. It’s still not completely clear if this is a good thing, though. That will be what we look at tomorrow.

Adjusting the Hitter/Pitcher Split (Part I)

No Comments
January 20th, 2010 by
Categories: Theory

One of fantasy baseball’s basic rules of thumb is the idea that you should spend about 70% of your auction dollars on hitters and the remaining 30% on building a pitching staff.

This is a useful guideline, but, of course, any rule will have exceptions. For leagues that start more pitchers (or fewer hitters), your budget will probably be slanted closer to 65/35 or even 60/40. Since the split is league dependent, the Price Guide does not presuppose any split but allocates dollars without regard to hitters and pitchers. (Typically, it will end up close to 70/30 anyway.)

I noticed last year, however, that the Price Guide tends to put a higher value on top-tier SP than typical auctions and drafts. Whereas a typical draft puts the first pitcher off the board late in Round 1 or early in Round 2, the Price Guide was suggesting pitchers from the start. That trend is continuing this year: Roy Halladay will probably be valued by the Price Guide as the #1 or #2 overall player in most leagues.

I’ve wrestled quite a bit with whether the Price Guide or the conventional wisdom is correct, and I’m still not completely sure of the answer. Regardless, I’ve decided there’s a good bit of value in at least allowing the Price Guide to create values that mimic real life.

So you will notice a new feature this year to enter your own hitter/pitcher split:

Setting a custom split with the Price Guide

This allows you to force the Price Guide to use a split that is realistic for the way your league really drafts. If your leaguemates will spend 72% of their money on hitters, you can enter a 72/28 split to get a better idea of how they will bid. If you’re in a draft, you may just need to gradually nudge the split in one direction until it matches what you expect to see.

Part II of this series will experiment with this feature for a real-world league, the experts’ Tout Wars Mixed League. In Part III, we’ll look at what the implications might be from letting your league dictate the values.

Prospectus Idol’s Fantasy Articles

4 Comments
June 2nd, 2009 by
Categories: Theory

You may have noticed that Baseball Prospectus is having a writing contest this summer, with the winner offered an extended writing gig with the company. The competition — Prospectus Idol — is styled after American Idol, complete with three judges and fan voting.

This week the contestants have written articles about fantasy baseball, and two of them have taken on the issue of fantasy player valuation: Brian Oakchunas submitted an article on a valuation system he calls VOFP — Value Over Fantasy Player. Tim Kniker’s article is for his Marginal Positional Value (MPV) methodology.

Now, there are plenty of fantasy valuation systems out there, including the Price Guide that you find on this site. In studying many of them, I’ve discovered that there are essentially three common aspects of all good systems:

  1. Converting the stat categories (HR, SB, etc.) to a common scale.
  2. Adjusting for position.
  3. Adjusting for the baseline.

Numbers 2 and 3 are related, and it’s possible to combine them into one step. (The Price Guide, for example, only makes one adjustment for the positional replacement level.)

Also, #3 isn’t required for straight drafts, although it can help give a sense of scale for how good or bad a certain player is compared to the benchmark (either average player or replacement level player). A replacement level baseline is essential for auction leagues, where everyone is valued compared to a $1 player.

So how do the two Prospectus Idol systems look?

Here’s how Tim explains MPV:

MPV is calculated by determining how a given player, if inserted into the baseline roster, changes the team’s final statistics in each category. [...] The MPV column is simply the sum of the percent increases in each of the scoring columns, since in this league each category counts equally.

Basically, to value a player, you are adding his stats to an average fantasy team and subtracting from that team the stats of an average player at that position. That gives a percent change in each category between the new team and the average team, which Tim sums to get a player’s total value.

So the MPV system handles all three steps in valuation. My main issue would be his choice of baseline (players projected for 200+ AB), which seems geared towards his 20-team league. As commenters on the article mention, people would probably want to find a more appropriate baseline if their league varied much from that format.

Brian’s VOFP is based on the amount of each category it takes to move up one spot in the National Fantasy Baseball Championship rankings (#1 above). His baseline (#3) is 40th place in each category (out of 390 people), which he estimates is the average finish needed to win the championship.

He shrugs off making a positional adjustment (#2), and I think that is a mistake. His comments give me the impression that he understands the idea of positional scarcity, and he’s expecting the user to either do that work on their own or rely on their intuition.

Brian correctly notes that there is insight in the unadjusted number… That’s why I think the best thing to do is to list both numbers — before and after the positional adjustment. That’s what the Price Guide does and also what Tim advocates in his MPV article.

My overall thoughts on MPV and VOFP:

1. For their specific environments, the two systems presented in the Prospectus Idol contest look to be fairly good at estimating player value (although VOFP’s lack of a positional adjustment is, in my mind, a big negative for that approach).

2. Both MPV and VOFP are described in very league-specific terms and would need to be modified for any league variations.

3. Neither presented anything that I would consider to be an improvement over existing valuation methods.

Dealing with Bench Players

8 Comments
March 9th, 2009 by
Categories: Strategy, Theory

As the fantasy draft season heats up, a common question I get is how to deal with bench spots when building dollar values in the Price Guide. My common answer to that query is, “I don’t know.”

By definition, a bench player is not contributing any stats to your team. That means, in theory at least, that a bench player is worth nothing.

In real life (real fantasy life, I guess) that isn’t the case. Players that you draft as starters will get hurt. Others will underperform. It is very likely that some of the guys that you draft for your bench will see a significant amount of time in your lineup, contributing stats toward your totals. Some bench players will have a non-zero dollar value, which is why owners will pay more than the minimum for them at the draft.

So how do we quantify the value of bench players? Here are a couple of tips for dealing with the issue:

Reserve some money
When you enter your league’s salary cap in the Price Guide, you might try lowering the amount to compensate for what owners spend on their bench. Suppose your league has a $260 cap that is also spent filling five bench spots. Maybe you only enter $250 into the Price Guide — allocating $2 per bench spot.

The average amount that teams spend on their bench players will vary depending on the league’s rules, so there’s no magic amount to allocate. Past drafts should give you a rough idea of how much you need to hold back, though.

Enter more starters
For a league with daily lineup changes, it’s common to keep a couple of extra SP on your bench, which you rotate in on days they are pitching.

If this is a widespread strategy in your league, then these players are essentially starters, and you can consider them as starters in the Price Guide. For a league that starts 4 SP with teams average 2 SP on their bench, you might just enter 6 SP into the Price Guide.

This also works for hitters, but teams aren’t able to rotate bench players in as consistently as they can for pitchers. If an average team in your league has three hitters on their bench, you might add one extra Utility starter to compensate for the combined contributions of those three bench players.

Inflate starter values
For leagues with large benches, you might notice that the deep draft pool makes teams spend a little more on starters. If this is the case for you, you might inflate the values of starters by a certain amount, say 10%. Just copy and paste the Price Guide results into a spreadsheet, and you should be able to tweak the prices until they make sense.

I really don’t like to advocate playing with the dollar values until they look the way you want them. When there are discrepancies between the projected values and what happens at auction, my preferred solution would be to identify the factors that cause the differences, so that the values are computed as accurately as possible. But the effect of bench players seems like a really difficult thing to quantify.

Does anyone else have any solutions for how to handle bench players?

« Older Entries