Mock Battle: Position Scarcity vs Best Actual Value
We are currently in the middle of round 12 in our early 12 team, standard 5×5, 1 catcher mock draft.
Devin Mesoraco was just taken at pick 137. I have him ranked almost 100 slots lower at 231 overall, and that is with a catcher position scarcity adjustment, so either way, I am considering this very early.
Up until this pick, only three other catchers were taken: Brian McCann (12.133), Yan Gomes (8.91), Jonathan Lucroy (8.88) and Buster Posey (3.35)
Below are my catcher rankings using Steamer’s projections (with a few manual adjustments) and Zach Sander’s FVARz approach to fantasy valuation (z-scoring each 5×5 category). Column 2 (zSUM) is the sum of the 5 categorical z-scores. Column 3 (PosAdj) is the zSUM adjusted for position scarcity.
There are 12 teams in the mock. Based on the number of teams, name and value, I’m assuming 15 catchers get drafted. Therfore the 16th catcher (Yasmani Grandal) is our “replacement catcher.” All catchers zSUM’s get adjusted/inflated by 4.52. Grandal’s adjusted value becomes 0.00. Buster Posey’s 3.46 jumps to 7.98. It’s this adjustment that knocks Posey’s value into the top 20 overall.
The position scarcity adjustment for Evan Gattis knocks him into the top 60. Right now, at #58, he is the most valuable asset left in my rankings. Justin Morneau is closely behind at #63. A couple of closers are still in the top 90. A couple of starters are still in the top 110 and then there are two offensive veterans I do not want to own in the top 110.
This means that until a cluster starting at #112 in my rankings, Evan Gattis is my best available option. His position adjusted value is 4.80 or he is worth about $16.00 if we are associated dollar values.
I am happy with my 1B, RP and SP and don’t want the two offensive veterans available before #112. I also don’t have a catcher yet. What do I do? Draft Gattis, right? Meh.
At this point, there are a number of catchers left that can provide similar overall value and/or power specifically: Salvador Perez, Yadier Molina, Matt Wieters, Wilin Rosario and Wilson Ramos to name a few. Because of these options, what I am looking at right now is still their raw z-sum values despite position scarcity. I want the best output (or soon specific counting stats like stolen bases).
Evan Gattis‘ zSUM is 0.28 relative to his 4.80 PosAdj value. Morneau is next up (not that I need him) and offers over three times the value at 3.00 even, but because he is at first base, his position adjustment only gets him to 4.63.
We are almost halfway through the draft. We only need one catcher (otherwise Gattis’ value would be even more inflated). At this point, I’m going with the better actual value over the position adjusted value until my options are depleted at those scarce positions. I am going Justin Morneau (even if my 1B and corner infield roster spot was already filled) over Gattis at catcher. If say, my 1B, CI and utility spot was already consumed, then I would probably go Gattis.
When you pull rankings from sites, you often just see an end ($) value. I recommend running the z-scores or categorical SGP values so that you can determine the best available value versus a position adjusted value, which whatever site you are on determines.
Daniel Schwartz contributes for RotoGraphs when he's not selling industry leading thermal packaging. You can follow him on twitter @RotoBanter
In a 12 team mixed league with only 1 catcher, why even do a positional adjustment? My home league is set-up this way and I never bother to adjust explicitly for positional scarcity for catchers. I just do my SGPs based on ZiPS/Steamer and go from there. Now doing this method ensures that you end up with the 10-12th ranked catcher, but who cares? Like you pointed out in your example Morneau is more valuable by raw numbers so I would take the (higher rated player) + (10-12 ranked cater at end of draft) over the (4th ranked catcher) + (whatever dumpster fire is available in the last round that isn’t a catcher)every time.
Why do you expect 16 catchers to get drafted when only 12 are required? Are 4 teams of experts really drafting an extra catcher?
personal rankings adjustment based on name value, rankings and value. I would draft gattis + another later in draft per Gattis’ OF eligibility posibility…sure.
Position scarcity adjustment still uber helpful in this draft and any (how to devalue pitchers relative to hitters even).
posey woulda been taken rounds later if it wasnt for the adjustment.
it’s always necessary…again, IMO. It’s about when it’s necessary.
@srpst23…if you do your SGPs based on your projections, you will never know how to value guys across positions.
I forgot to mention that my home league does not have a bench. So in my league no one should ever be drafting two catchers. I could definitely see how some would use a bench spot to rotate catchers on off days or use a Gattis as a plug in for the outfield. But I still would rather have a Grandel(your replacement level) + Morneau over Gattis + last round Flier.
I’ve done the positional adjustments in the past and in a 12 team league I don’t really think that it’s useful. No position is really shallow enough to warrant it. Some may seem like it, but after everything is calculated from the projections it usually isn’t. If you are using a good projection system for your SGPS and dollar values, you will still end up with a decent player at every position because a $4 true value second baseman is what maybe $5-6 after positional adjustment. So instead of thinking that I got a bargain because I got a 6$ 2B for $4 I just think that I paid what he was worth. I prefer to spend my $$ on the actual production at the top end of my draft, not positional scarcity. There are always going to be decent options if you really stick to your projections/$ values because you won’t get caught in a bidding war and spend $25 on the “hot sleeper” or some other overvalued asset. My teams tend to be rather boring, but also usually pretty good because I end up with a lot of the unsexy undervalued veteran players,
Is there a reason Travis D’Arnaud is in a green box on the spreadsheet? Perhaps I should be even more excited about him this year than I already am?!? I picked him up for nothing late last season on my Dynasty team.
Oh I am an idiot, I had clicked his name there on my screen and made the green box myself. Anyways….D’Arnaud!
rookie move bro. clean it up.
I think there is something wrong with your Zscores, McCann should not be a -.23…
if you average the 5 z-score columns in gray that sums up to -0.23:
1.1 -0.36 0.45 -0.95 -0.47
yeah I think you are using fewer players to calculate the z-scores than I do. Not necessarily incorrect but to me, saying McCann is a below average player seems wrong. Or, that Lucroy is just average..
He had a 93 wRC+, with 23 HR, 57 R, 75 RBI, 0 SB, and a 0.231 Avg. That seems pretty average to slightly below average to me. The only reason he’s valued higher is because he is a catcher, and his adjusted Zscore is 4.29, which makes him much more valuable if you believe in positional scarcity.
Well, fundamentally that is only average if you are only looking at, say 230 hitters or so. If you expand it to 300 hitters it’s well above average. Also we are talking about his 2015 Steamer projection, not his 2014 performance, although it’s only a bit different.
Don’t sum standard deviations. Ever.
FWIW, the avg. and SD for each category is taken on all hitters with 450+ PA i believe it was.
If you believe in the projections, and don’t have an issue with avg, then Gattis is the pick. If you don’t believe in the projections then it’s Morneau. The person who wants to win takes Gattis. The difference between him and the other catchers is vast. Morneau type 1Bs are not as hard to find. That is if you believe the Gattis projection.
Your decision may be correct, but your justification is not. Your unadjusted (for position) projections have Gattis significantly superior to Rosario, Molina etc. So if you think they will provide similar value, you’re not ignoring scarcity at all but rather doubting your choice of projection.
Is catcher really that shallow? I would be comfortable with either Grandal or D’Arnaud as my fantasy team catcher.
I think there’s an empirical answer to this. Imagine 2 teams drafting in a snake draft. One team drafts purely based on overall values and only pays attention to positional scarcity, the other team drafts purely based on positionally adjusted values and ignores non-positionally adjusted values. Assume that the projection system used is a perfect predictor of season performance. Who wins the offensive side of league if its a standard 5X5 roto league? We can punch the projected stats in and have a precise answer to this. Do you have your entire spreadsheet of values somewhere? I would be very interested which team wins that hypothetical 2 team league, and I think that clearly answers the question of whether you should positionally adjust your values.
Rd 1- Player A takes the best positionally adjusted player while player B takes the best overall player.
Rd 2- Player A takes the best positionally adjusted player at a position he has available while player B takes the best overall player at a position he has available.
Rd 3 through however many roster spots you’re filling- Repeat process of Rd 2.
Then stick their projected Stats into a team ranking of your 5 categories and see which team gets the most points. I’m betting on Player B, but I would really love to know the answer to this. If you have your overall rankings and projections somewhere it would be easy to do.
Doesn’t work that way, since a 2 team league would have a vastly different replacement level for each position. You’d have to do multiple simulations in a 12 team league, based on multiple draft strategies of the other 10 teams. And I have a hard time believing the strategy that utilizes less important information would be superior. You’d basically always be drafting one way along the defensive spectrum, and that could easily be exploited.
Hmm, take 6 “a” teams and 6 “b” teams. “a” teams use strategy a and “b” teams use strategy b. Run simulations. Sometimes a’s go first, sometimes b’s go first, average out. Pick winner.
Bounty, I really like that idea and want to do this. Would love to see Dan’s spreadsheet for the non-catchers.
For a fictional 2-team example, it’s pretty trivial to prove mathematically in a simple case. Let’s say we have two positions, Goods and Bads, each with two players (payoffs shown in parentheses):
Goods: A(+10), B(+8), ..
Bads: C(+6), D(0), …
If you go first in this draft, who should you pick to win?
Answer: Player C, in the Bads position
While C’s +6 points is 4 less than than A could provide, you can always get B on your next turn. Basically, the winning strategy gives you:
C(+6) + B(+8) = 14 points
Loser gets:
A(+10) + D(+0) = 10 points
I mean, you can still win if you pick A, but it requires the other drafter to also be dumb about positions and goes with a max-value strategy.
I should also note that if you are sure your opponent is dumb (doesn’t consider positions in such a game) and that there’s some randomness in those payoffs, you might still want to go for max value because you could get the situation:
A(+10) + C(+6) = 16
vs.
B(+8) + D(0) = 8
But that’s taking a risk about the strategies people are using.
If its a standard snake draft, the second person to draft will always win in your little game, I don’t see what this has to do with the question. If player A is taken first then the second drafter will get picks #2 and #3 and will take Players B and C and win 14-10, if Player B is taken first then the second drafter will take Player A and C and win 16-8. There is no way to win drafting first in your little game. I don’t care about draft position, I don’t even play snake drafts, I play auction drafts, but the question is, if we positionally adjust, do we give up the value in the best players and lose the league?
I just recalculated my z-scores for a positional adjustment, Dee Gordon stole 64 bases, if you draft him as a SS, those 64 stolen bases had a z-score of 2.782. At catcher, Carlos Ruiz was tied for the NL lead in Stolen bases at 4. His Z-score was 2.184 for the stolen base category. Are people seriously telling me that 4 stolen bases from Ruiz is only slightly less valuable than 64 stolen bases from Gordon because Ruiz is a catcher?
LOL no one intelligent uses position-specific standard deviations in calculating z-scores. Try again.
Yeah it would probably work better to just have a ‘average sum z score’ for each position in order to calculate the positional adjustment – in order to counteract anomalies for certain categories at each position.
CM2, then what the hell did Dan Schwartz do? I agree with you, it would be ridiculous, but I don’t see what the positional adjustment is then, and I read through the entire z-score series and I know they were advocating for position-specific z-scores in those posts. I agree with you, its a crazy thing to do, but as far as I can tell that’s what Schwartz and anyone else exactly following the rotographs z-score method is doing. Where does the positional adjustment come from if not from z-scores?
Why are you asking me what he did? Read his article and the links he cited.
In a nutshell, position defines the replacement level. But a marginal stolen base from a C isn’t worth more than a marginal stolen base from a SS because catchers have a lower standard deviation. Positional adjustment is a constant translation based on the replacement level and that is all. Look at the table above for catchers. How many positive values of zSB do you see? That should have clued you in that your calculations were ridiculous.
Here you go. Mere sentences into Sanders’ article:
“After comparing players to the entire pool, their zWAA is compiled and compared to their position. After the positional adjustment for replacement level, a players’ final value (zWAR) is produced.”
I’m not making positional adjustments at all, and I think the z-score method they’re using of all players with 400+ plate appearances, perverts the idea of “replacement level.” I don’t really see why the hostile tone. I’ve read the articles, I’m trying to figure out what they’re actually doing. Not sure why the hostile tone, as far as I can tell you and I agree that z-scores shouldn’t be calculated for each position. I do a Z-score comparison for the entire player pool, then obviously I need a catcher, so I find a catcher, generally a shitty one because I don’t think mediocre catchers have much value.
Maybe you can enlighten me, what does “compiled and compared to their position” mean mathematically. If you want to positionally adjust the way they’re doing it, I was under the distinct impression from that series of articles that they were building z-scores within each position. If that’s not what they’re doing then I don’t understand what they ARE doing to positionally adjust.
For what its worth you’re referencing a corrective piece to an earlier version in which he absolutely was building z-scores for each position, but I guess I’m not following what he’s actually doing now in the positional adjustment.
Ok, so they’re not doing z-scores by position anymore, what exactly IS the positional adjustment now then? I don’t know what “compared” means in terms of practical implementation.
Here’s my best guess, which is probably wrong. “Compared” means…
[(average z-score for catchers)-0] + (z-score for each catcher)? Then repeat for every position? Is that what he’s saying?
@Corey: That’s probably about right (or would give the same inference). Basically, the normal distribution has two parameters: location (mean) and scale (standard deviation).
First, they calculate the mean and standard deviation for all the players for a category (e.g., runs). Then they subtract the mean from all the scores and divide by the deviation. After they have done that, they usually average/sum across the categories for each player’s z-scores. Then, they rescale the average z-score for each position by taking the average of it and bumping everyone up (or down) so the mean of player value is worth zero for each position.
For any of these things, you don’t need to use the mean as the location parameter. You could use the worst player’s value, for example, it really wouldn’t make a difference in your decision-making based on it (differences between values would remain the same).
It should also be noted that raw z-scores for rate stats are incredibly wrong because they seldom consider the weight of those stats (e.g., IP, AB, or PA). For their z-scores to be useful at all, they need to be transformed into a form that represents the amount of deviation they would induce from some benchmark location (e.g., overall group ERA, replacement level ERA, whatever).
However, z-scores are hardly the only (or the best) game in town. If you know more about the distribution of each category, in terms of how it leads to points in the league (e.g., typical cutoffs for ordinal rankings), you can consider each player’s contribution as a share toward those points. This uses more info, so it’s almost uniformly better if you can do it.
@Corey, re: “If its a standard snake draft, the second person to draft will always win in your little game, I don’t see what this has to do with the question.”
I don’t think you are that good at this game. Assuming you need a player for both spots, Player 1, if smart, should always win, even in a snake draft:
P1: C (6)
P2: A (10)
P2: D (0)
P1: B (8)
P1: C (6) + B (8) = 14
P2: A (10) + D(0) = 10
And the relevance of that example is that it is the most minimal example of a draft you could have. Two positions, two drafters, two players. And, even in that minimal case, we can *clearly* see that considering relative position strength matters. Based on this mathematically-provable game, you can extend it to pretty much any snake draft and can apply similar insights to auctions.
@Corey: Also, the winning strategy is to pick neither A nor B first. You pick C. It proves EXACTLY what you were asking:
“If we positionally adjust, do we give up the value in the best players and lose the league?”
The answer is no. Value is only as good as its improvement over replacement. In this case C has a replacement value of +6 and A has a replacement value of +2. So he’s the first guy you should draft, even though he will not produce the highest stats.
Z, thanks for your responses. I’m sorry, but I really don’t follow the relationship between the mini-game and the question because we’re arguing over the values, in your mini-game whoever takes the 0 value player loses, so if you pick second you always win because you can avoid the 0 value player who has a tremendous marginal effect when drafting out of a limited pool of 4 players. I think you’ve made the point that others made that my initial idea was probably too simple to reflect anything except a league that would be too boring to bother playing. You just had player 2 take the 0 value player with his 1st pick of the second round which there is no reason he would do. In that game, there is no reason that player two would want to take the 0 value player, so purely by his draft position he’s assured victory no matter who player 1 takes, there is no winning strategy for player 1 except to take the best player, I don’t see why you had player 2 take the 0 value player when he can avoid it.
To me, I don’t do positional adjustments and if I’m looking at my roster and I haven’t filled a particular position and its obvious that I don’t want anyone left except a particular player, I’m willing to slightly overpay for that last reasonable player, then I don’t really care who I grab for $1 after that because they’re all terrible. Snake drafts are a little different, and I don’t do snake drafts.
The influence of weight in the z-score method is a definite problem, what the Sanders system advocates is a “weighted OPS” and a “weighted Batting Average”, same with pitchers for ERA, WHIP, K/9. I’m not completely sold on that, but it does seem to correct at least some of the small sample success bias.
I would love to see a series of articles on fangraphs like Sanders’ discussing other ways to value players specifically for your league’s context for custom values. I’ve seen some discussions of that in comment threads, but I found the Sanders piece while I was trying to figure out how to do individualized valuations off a hunch (I think totally right) that values you find in magazines aren’t matching my league structure. I think the z-score method has worked relatively well, but I’m really interested in other methods. All I can think for another method is take every team’s roster and subtract out each player on it, and see how many points they were worth, I’m not sure how you would extrapolate that to the next season though, how do you break that back to a projection so you can use it for the next season? Mock draft a league based on the projection then pull them out like we did post-season? That seems like about 75,000 X as much work as the z-score method and would also create bias towards team composition.
I would have taken Gattis. He will play LF and likely get more ABs than any actual catcher. A power guy with catcher eligibility will be a great value.
I did a similar method last year (z-Scores), with mixed results. A few issues that I found:
1. Normalization Issues – One very important thing is to only do positional adjustments based on the number of people who are expected to be drafted at a position. Due to some positions being more generally powerful (and hence more attractive bench players), this is actually somewhat non-trivial. I find it better to normalize the scores based just on the guys I am SURE will be drafted, and then apply this Z distribution. If you don’t do this, guys like Gattis look downright godly, because the bottom of the catcher bin is terrifying to behold.
2. SB Bunching – Raw Z scores tend to undervalue the importance of high-rate base stealers with guaranteed jobs. This is mainly because the distribution is messed up by tons of assumed stolen bases from light-hitting guys that are projected to receive playing time but might not. When I used this method, I fell quite short on SB compared to not using it.
3. R/RBI – This method, however, is pretty great for grabbing harder to conceptualize quantities like runs and RBI’s (e.g., the actual advantage of having a catcher who knocks more runs in).
4. Pitcher Playing Time – Pitcher playing time varies due to a lot of factors not included in the projections (e.g., health, role assigned during spring training). In general, because of that volatility for the younger, high-yield pitchers, every time I have used a highly quantitative method for drafting, old farts are overrated. The projections tend to give them too much time (i.e., can’t predict injuries) and the young guys too little time (e.g., can’t predict full-time role status). Unless someone knows a projection system that follows the news, I don’t know a great way to avoid this.
I should clarify my first point a bit, at least as far as how I normalize:
1. Normalize by all hitters or pitchers that you think will probably be drafted (e.g., will have some bench players)
2. Then you do a positional adjustment based on the guys that you think will SURELY be drafted at each spot (e.g., don’t bump up Posey’s positional value because of catchers that might not even be drafted)
I think I disagree with point 2 based, but agree on main points from 1, which is why it’s by league/team size, best value/zsum and “name.” In other rankings it wasn’t perefct order and I’d include a zsum like bogaerts if the projection system said he was below standard replacement level simply by zsum. For this, it was pretty on for the most part (steamer had most draftable “names” above by glance so I kept it completely objective. I think OF position might be the only position where we would have wanted to pull up names from below replacement level, but on phone and can’t check now. It’s an art: imo it’s zsum first, name and #per position second.
@DS: The problem with including guys below the bar of “certain draftees” is twofold:
1. Structural variance: These guys often have tons of playing time variance. A possible call-up who is most likely to play either 10 games or 100 games gets averaged to 60 games (which he is least likely to have). It’s like Schrodinger’s cat: his major-league job is alive or dead, but a probability of 0.5 does not make it half-alive. You can rescale everyone’s stats to a static # of games, but this overvalues platoon players, injury risks, etc. In short, leaving raw projections for these guys gives too low of an average level (the ones who play will produce more than stated) and too high of a variance. Worse, adjusting them to a full season makes the average performance too high and too low of a variance.
2. Including on-the-cusp guys in your z-norm will tend to make their replacement level seem much worse than it actually is (for the reasons stated in #1). This is particularly true for outfield, which has 3x as many players as other positions and platoons are common. Across the season, the following things happen:
1. Players’ playing time/quality changes (injuries, call-ups/breakouts, platoons, trades, lineup position)
2. These changes are known and are structural (e.g., make a persistent change in a player’s future projections)
3. Positions with more players will have more players in churn (in absolute # of players) at any given point in time.
4. Hence, the effective replacement level of those positions (typically OF and SP) tends to be higher than expected for the guys who are uncertain to be drafted.
Or, short issue: Since there are so many OF/SP, we will see more of them (in absolute numbers) seize big playing time and pick up big numbers. They can then be played until their next structural shift.
One could solve this, I suppose, by just merging the fringe players (e.g., “platoon” their projections until you make a full season), but I tend to favor just leaving them out entirely.
I have a real problem with the cited form of calculating ‘position scarcity’. It greatly overvalues mediocre players and pretends that projected values are more accurate than they really are.
The ‘replacement’ or marginal catcher is NOT the 13th or best undrafted or whatever. The replacement catcher (or any other position) is the best catchers who remain available when one is making a specific draft pick. Posey’s positional scarcity is not valued relative to Yasmani Grandal. It is his value relative to Gattis/Lucroy/McCann. And their positional scarcity is valued relative to d’Arnaud/Martin/Perez.
There is no scarcity of actual position players at any position (except in a 10 team NL East only league). The difference in value between Zunino/Molina/Grandal is basically nothing – well within any likely projection error. And since one of them is projected to be available on the waiver wire (for the cost of a mistaken draft pick), it doesn’t matter much at all in general terms which one is, basically, a last round pick.
Yet every year, I see players overdrafting the slew of mediocre basically-replaceable-level 2B/SS/C (and far worse than replacement level fantasy batters) because ‘position scarcity’. And even more laughably, filling up their bench with that stuff. Not that I really mind that much because those are the players who allow me to win most leagues I play in.
Individual players certainly do have ‘position scarcity’ because of the slots you can put them in score. But that scarcity only applies to those individuals – not to every player with that position eligibility
I would counter-argue a bit that it matters in terms of major dropoffs. You have 162 games to expend at the catcher slot, for example. If there is a dropoff of 1 STDev at that position, and no other comparable dropoff for other positions, you may be well-served to grab that last elite catcher.
The issue that you’re talking about comes when you do positional adjustments that basically rank each position ordinally, rather than as assets that produce continuous values (e.g., #6 will produce 10 runs more than #7). Often, you see people stupidly grabbing a SS because they’re “scarce” (e.g., among the last of the top 10 SS). However, the dropoff between the 10th SS and the 16th is often pretty laughable. As you say, well within projection error. However, sometimes there is a huge dropoff where you just simply will never get much value out of a position after that point, period.
And by 1 standard deviation, I mean one deviation scaled based on all batters’ projections.
Yes agree with z here. Position scarcity important so you can understand clusters within position …if Gattis is standard devs from rest of group, he may not be best avail but could be best option. That’s what I meant in article by depends on options left within position…”z” better worded it
And by group I mean other catchers!
I agree with the major dropoffs. And the method can identify those major dropoffs, help quantify them, and help assess which positions have a major dropoff when other positions might not at that particular draft slot or $ auction value.
But generally fantasy owners will still feel compelled to plug in a position scarcity number for far too many players simply because a number is available/calculable.
Using the example of SS, I’ve found that the true result of ‘position scarcity’ is that only 5 or so are worth spending any drafting effort thinking about. They have true position scarcity. The rest are a big not-worth-thinking-about. Let other players chase those guys a few rounds earlier than they should or spend their money on them. If you don’t end up with one of those 5, then yeah you are in a hole. But so are 6 other owners – few of whom will believe they are actually in a hole because they wasted money on the mediocre ‘7th best SS’.
Thank you for providing a perfect example of why Z-scores are not a good way to rank players. You need a system tells you not what the dispersion between players is but rather a system that tells you how many points you actually gain from that player. SGP is one approach that people are familiar with–I use a different take on it where I determine the number of each stat that will be counted over the course of the season and rank on player shares of that stat.
Do it that way and positional scarcity adjusts itself due to share differences by position in generation of statistics.
Z-scores won’t get you where you want to go.
Technically z-scores can get the same inferences if you’re normalizing them right, but that is definitely a superior method if you know how the league structure tends to pans out in terms of how stats match up with rankings.
Yeah SGP is better IF you know league structure, history, etc. I like the z approach for general here you go rankings. Also can trends effect SGP more than z’s? Maybe same diff?
@Blue – That method sounds interesting. How do you rank player shares for the rate categories?
Z-scores are a better way to rank players IMO. Z-scores opine on the distribution of a category, SGP does not. Example. Take a world where a hitter (in a universe of 10 hitters) represents 20% of that category:
Scenario A: 10, 10, 10, 10, 10, 0, 0, 0, 0, 0
Scenario B: 10, 5, 5, 5, 5, 5, 5, 5, 5, 0
Scenario C: 10, 8, 7, 5, 5, 5, 4, 3, 2, 1
SGP would say the player who scores 10 in any of the three scenarios has the same value. But Z-scores would say that a player who scores 10 in Scenario B is most valuable.
Separately, SGP would value the player who scores 0 in Scenario A the same as Scenario B. If I was forced into taking a player who scores 0, I’d rather have it be in Scenario A versus B. SGP does not have an opinion otherwise, whereas Z-scores would differentiate between the two.