SubscribeI imagine that most people are already aware of the concept of tiers — the idea of placing players into groups with similarly valued players.
Today, I’m going to look at what I consider to be the benefits of using tiers for your draft preparation. I’ll address some of the problems with tiering tomorrow.
So what’s to like about tiers? Well…
Tiers recognize that projections are far from exact.
In fantasy, it doesn’t really matter much if your projections have Alex Rodriguez valued at $43.48 and David Wright at $42.21. At the end of the year, either one has essentially the same probability of outearning the other.
Ron Shandler has noted that even the best projections will only fall within +/- $2 of what a player earns about 46% of the time for batters and 37% of the time for pitchers. With that margin of error, it’s not a big deal to go to $45 on David Wright. It’s not the end of the world if you stop at $40 for A-Rod.
With tiers, we don’t make a big deal about that $1 difference in their projections. We put A-Rod and Wright in the same tier. We know there are pros and cons with either one, but we also recognize that they are very similar in value.
Tiering recognizes that the important thing is realizing that both players should be valued somewhere around $40+.
Tiers show where there are gaps in value.
So we have put A-Rod and Wright in the $40+ Tier. What is the next tier of 3B? According to CHONE, it’s Aramis Ramirez and Garrett Atkins, both at $19. (See note on Longoria below.)
That’s less than half of the value of our top tier players! In draft terms, we’re dropping from early first round players to somewhere in the fourth round. Below these two $19 players is another $7 drop down to Chipper Jones at $12.
Mind the gap!
These gaps between tiers is especially important information in a draft: Let’s suppose it’s your pick, and you are choosing between Atkins and Brian Roberts. You have Roberts ranked marginally higher than Atkins. Your 2B rankings have Brandon Phillips, Dan Uggla, and Robinson Cano ranked just below Roberts — all in the same tier. Atkins is the only remaining player in his tier at 3B.
Although Roberts is projected to be worth more, in this situation, you might consider taking Atkins over him. The values are close, and there are other 2B options in the same tier as Roberts.
All else being equal, you should prefer to pick a player at the bottom of one tier over a player at the top of another.
The main things to take away from tiers are that they can help you to mentally group similar players, and they can show you where gaps lie between players at each position.
But are there situations where tiers don’t work? That will be our topic tomorrow.
*”What about Evan Longoria?” you ask. “He’s been going for $25-30 in auctions.” Well, for this example, I’m just using the CHONE projections, and CHONE puts Longoria at $5 in a standard mixed league. “Five dollars?!”
Now, I’m not saying that I agree or disagree with that projection, but it does give me pause… There could be some very disappointed Longoria fans this year.
Related posts:
I’ll be very curious to read your continued thoughts about Tiering – as I study the Price Guide, I keep wondering how a tiering concept might be built in – in particular, whether there’s a role for conditioning your positional adjustments (based on a VORP-type calculation) with a secondary adjustment for a player’s value over the NEXT player or players at his position. Because the quality distribution is not particularly even, the Price Guide’s uniform positional adjustment may be suboptimal.
Anyway, keep up the great work – excited to see the ‘composites’ up.
the above point is something I considered as well.
instead of taking the average and stdev of all players stats and giving each player standard score, what if you took the average and stdev of players at each position and then gave scores based on that?
ie Arod’s standard score would be based off only other 3b. would this give you a better idea of scarcity/value within a position?
@Homer: I don’t know if there’s anything wrong with that, but I don’t know if it helps, either.
Looking at individual positions, you are starting with a much smaller sample size, which means the results could easily be skewed based on one or two players.
Mike Podhorzer has mentioned an approach similar to this here. It ends up suggesting (for example) that HR are extra-valuable from 2B/SS, while SB are more helpful from other positions.
@Homer – that approach doesn’t work. It doesn’t allow you to compare between positions at all.
Think about your final standings, you don’t care about where you get your 250 HR, just as long as you get 250. A HR is worth the same regardless of position. A system like this doesn’t work because it can’t compare between positions. What if you had a utility/DH spot? Then you end up with player A having more value than player B at his position, but at DH player B has more value than player A. This is clearly not right.
To prove this, I ran the 2008 stats for my NL only league.
Pujols is worth $38 as a 1B, but he’s worth $41 as a utility player, and Berkman is worth $39 as a 1B but he’s only worth $35 in the utility slot.
Things are even more screwed up if you have an SP and an RP slot: the top two relief pitchers are Hong Chih Kuo and Carlos Marmol! Low ERA/High K/High Win relievers are absurdly valuable under this system. A win from a reliever isn’t worth more than a win from a starter in the final standings, but doing positional z-scores it’s almost 2.5 times as valuable!
I never understood the attraction of tiers. To me, it just seems like a shortcut to value/rank players. Why do I need tiers if I already have calculated actual dollar values? The dollar values themselves show me where the drop offs are and I still would like to know who I value more between Wright and A-Rod, even if they are only $1 apart.
I think it’s mostly for people who don’t know how to or don’t want to calculate dollar values, since it requires less work, but it’s obviously much less precise.
The example of taking Atkins before those 2nd basemen based on tiers is EXACTLY what I do. Although I technically just draft with my spreadsheet of players in descending order of dollar value, I color code each position which in essence illustrates the tiers. If there is a cluster of OFers, for example, at the top of my sheet for my next pick, with a 3B in between them and no others for another 40 players, then I go with the 3B even though a couple of OFers I valued a little higher.
@MikeP
I often color code or the like as well, but the fact that in circumstances like the Atkins one we thus stray from our ‘precise’ rankings illustrates the problem – I would think a slight correction to boost the computed $ value of the last 3B in a tier (or to lower the highest in a tier) would improve the situation (i.e., one wouldn’t be stuck as often having to decide whether to ‘go with your rankings’ or a vaguer tiered approach).
In an actual (snake) draft, where appropriately reacting to position runs is a crucial skill, this probably isn’t important – I was just thinking that for theoretical purposes (like the ‘Retrodrafting’ that is used for testing the Price Guide against other rankings) a ‘tiering’ correction might add value.
The correct thing to do is to readjust the value of each player based on who is already taken in the draft. Thus if you have 3 top-level SSs and then a huge drop off, the correct dynamic valuation system would give a bump to the 3rd SS that’s left – but not until the other 2 SSs are taken, since you’re already accounting for position scarcity in your rankings. It’s just that the position has now gotten shallower.
How would such a dynamic adjustment be calculated? The replacement level hasn’t changed, and overall STDEVs would not have changed in a position specific way. Wouldn’t you need positional STDEVs or something (and I see the flaws noted above) to show Rollins’s ‘increase in value’. (I have a feeling I’m missing something obvious).
If it is calculable, I would think we should be able to calculate it in advance, since we ‘know’ that the top 2 SS’s will be gone when it comes time to assess the 3rd.
That’s assuming that the players come off the board in the right order. If Rollins is drafted before Reyes, you’d want a different modifier applied to Reyes’s value than Rollins’s value, right?
I’m not sure how to do it though, as the method I was thinking of is clearly wrong when I think about it more, so I’m not sure that doing an adjustment like this is feasible or even correct.
I posted this on your last one for someone else but wanted to hear your thoughts on this quote (from rookies and cream, member of cafe)
Essentially, you are calculating z-scores for each category and summing everything up to get a total ranking. The problem is that the z-score method assumes that all categories are normally distributed. If categories are normally distributed, you could then actually convert z-scores to percentiles. For example, a z-score of +2 is at the 98th percentile. However, some categories (e.g., stolen bases and saves) are clearly not normally distributed. Thus, players with very high totals in these categories (i.e., outliers) are going to have total scores that are inflated. Do you correct for this inflation?
Also I saw you added composite stats which is really cool……not sure what you use though, was it the one I told you about at the cafe or is it like marcels and chone combined or something?
Yeah, some stats aren’t normally distributed.
Some are, though. R, HR, RBI, and especially rate stats have a pretty nice bell curve distribution.
Saves, SB, W, and K don’t, though.
The thing is, you’re not really using standard deviation to compare to a normal distribution. The way this z-score system works, someone 3 standard deviations above the mean simply gets 3 times the value for that stat as someone 1 standard deviation above the mean, as you’d expect. Someone with 30 HR gets a value of 30/SD, someone with 10 gets a value of 10/SD, so the 30 home runs is simply worth 3 times as much as the 10.
You are going to still overvalue the non-normally distributed stats a bit, but it still does a pretty good job of estimating the value of one stat w.r.t another.
@Mike Podhorzer:
“I never understood the attraction of tiers. To me, it just seems like a shortcut to value/rank players. Why do I need tiers if I already have calculated actual dollar values?”
I agree that they don’t provide any new information for an auction. I think the value is for drafters, who might not usually bother with dollar values.
Basically, if you are in an auction OR a draft, go ahead and build dollar values. They will tell you much more than a list of ranked players.