Prediction season is over, and the regular season has begun, which means it’s time for the second instalment of our win total prediction analysis. Last year, I gathered all of the win predictions from Effectively Wild podcast preview guests and our own preview writers, as well as PECOTA and Fangraphs projected win totals, and compared them to each other, then compiled projected standings tables and reviewed the results of a listeners’ PECOTA over/under poll. At the end of the season, I compared all the totals to the actual standings, with the EW guests emerging as comfortably the most accurate over PECOTA, Fangraphs and some people from a baseball site that shall remain nameless. Can they retain their title?

A few notes on the methodology: if two people were involved with a preview and therefore provided two separate win totals, I took the average of the two. Five BttP writers also didn’t include a prediction within their preview, so I solicited their total via Facebook instead. Despite the strict interrogation by our intrepid hosts, some podcast guests were a touch evasive in giving a single win total (saying 86 or 87 wins, for instance), so in these cases I simply used whichever number they said last, as technically their most recent prediction. For the listeners’ PECOTA poll, I asked respondents to choose whether they thought the team’s win total would be over, under or the same as each of the projected PECOTA totals, then also asked them to predict each division winner, the wild cards, the teams most likely to over- and underperform their projection, and whether their team’s fanbase thought PECOTA hated their team.

On the projection system side, EW Facebook group member Peter Quadrino very helpfully put together a spreadsheet that listed the win totals from multiple projections. I took the original PECOTA totals (used for the listener poll) directly from the BP projected standings page, but I’ve used his sheet for the more recent PECOTA totals and the Clay Davenport projections. As Fangraphs projected totals are based on a combination of ZiPS and Steamer with their own playing time modifications, I went straight to the FG projected standings page rather than include those two systems separately. With all that said, let’s get started with the level of optimism.

Level of Optimism

In what seems to be a common theme of human predictions, both of our sets were more optimistic than any projection system by a clear margin. Against 2430 total wins in any full 162-game MLB season, the BP Annual essay writers came in at almost exactly the same total as 2015, topping it by two with 2504 total wins from their combined predictions to be 74 wins on the optimistic side. BttP staff collectively predicted 2498 total wins, 68 higher than possible and 6 more than last year’s total. The graph below shows all 5 sets of totals in relation to each other; any of the sets can be removed by using the menu on the left hand side if you would like to compare a subset of these totals, or simply see one set.

It isn’t just that 3 or 4 predictions significantly skewed the results in favour of the humans being overly optimistic; both EW guests and BttP writers were pretty optimistic across the board against the projection systems this year. Among EW guest predictions, 20 out of 30 were higher than PECOTA’s total, 23 exceeded Fangraphs’ and 18 were over Davenport’s. Twenty BttP picks also went higher than PECOTA (albeit not the same 20) and 22 topped Fangraphs, with 19 exceeding Davenport. However, the most optimistic total for any individual team almost went to a projection system this year, as Davenport’s 100-win total for the Cubs was only pipped during the penultimate EW team preview, when Mark Armour and Dan Levitt combined to give the Cubs 102 wins. Davenport and Fangraphs also tied for the most pessimistic total, putting the Phillies at 64 wins.

On EW, the guests came in particularly high over Davenport – which seems to be a system prone to predicting a wider range than the others – a few times, including another Royals outlier (Sam Mellinger’s 87, +17), some relative Padres optimism from Michael Clair (79, +11), and Pete Beatty’s 91 for Cleveland (+9). Perhaps most surprising were Claire McNear’s 88-win prediction for Oakland, 12 higher than PECOTA and 8 over Fangraphs, and Holly Wendt’s Phillies love, with 75 topping both Fangraphs and Davenport by 11 and PECOTA by 9. Most of BttP’s optimistic predictions were still within a few wins of the systems, with the significant outliers coming from (unsurprisingly) the Royals, who Eric Roseberry picked to win 90 in the face of overwhelming projection insistence on a total in the 70s, and the Cardinals, with notorious St. Louis homer Alex Crisafulli pegging them for 95 wins, double-digits higher than all systems and even 6 over Dayn Perry’s prediction.

The differences weren’t anywhere near as extreme on the low side, but there were a few 7-win differences: on the BttP side, Tyler Baber’s 73-win Marlins prediction vs Fangraphs and the 80-win total for the Mariners from yours truly vs Davenport, while for EW David Roth called the Angels a 77-win team, agreeing with PECOTA but not Davenport. There were also two 9-win unders against the systems: Andrew Patrick’s 75 for the A’s was in sharp disagreement with Davenport’s 84, but the chief pessimist title has to go to Russell Carleton, who picked the Reds to win 65, putting him 9 below PECOTA and 7 below Fangraphs, as well as ten lower than BttP’s Barry Gilpin. Four BttP and EW predictions also matched exactly, for the Diamondbacks, Mets, Nationals and Orioles. Perhaps we listen a little too closely (or they read us a little too often).

Divisional Standings

Once again, although it doesn’t technically make sense mathematically for the predictions, I’ve compiled a set of sortable standings tables to show the differences between each set on a divisional level. The listener responses to the PECOTA poll have also been reviewed in relation to these standings.

AL East

TeamPEC Feb 17PEC Mar 27FGDavBttPEssay
Red Sox888788868984
Blue Jays868684888887
Yankees858482818584
Rays918781828385
Orioles727479778181

The AL East is still a bit of a mess to sort out this year, but not on the scale of 2015. All of the sets have either the Red Sox or Blue Jays in first place, although PECOTA still has the Rays tied with Boston for first after dialing back its early preseason optimism by a few wins. With that reduction in the Rays total, no team is projected to win more than 90, and PECOTA aside, even Baltimore, last in every set, is considered more or less a .500 team across the board. The listeners took their strongest stance (yes, even more than the Royals) on that early Rays projection, at just over 90% taking the under, and seemed to favour the Blue Jays (59.72% over at 86 wins) to the Red Sox (50% under at 88), with the Yankees responses fairly evenly split. That was backed up by a 33-29 win for the Jays over Boston amongst the divisional picks, with a handful of votes going to the Rays and Yankees, and the Rays being the most popular pick to under-perform their PECOTA projection. There was also a fairly heavy support for the over on Baltimore at 72 (70.83%), a stance that looks somewhat vindicated in the early going, although not a single listener picked them for a playoff spot, making them the only team in the AL not to be selected.

AL Central

TeamPEC Feb 17PEC Mar 27FGDavBttPEssay
Indians929286829091
Royals767577709087
Tigers797981818580
White Sox828480888583
Twins797878807683

The Royals screw up the Central somewhat, as all the systems hate them and the writers love them. Elsewhere, there’s pretty broad agreement about this division, with Cleveland in or close to first place, and the Tigers and Twins hovering around .500. Davenport – not for the last time – takes a different view by preferring Chicago to Cleveland in the biggest non-Royals outlier, while BttP previewers Mark Sands and AD clearly believe in the Tigers more than the rest. The listeners, like the writers, vehemently disagreed on Kansas City, with close to 89% taking the over on 76 wins, and almost as many (83.33%) picking the under on 92 for Cleveland. KC were also the overwhelming favourites to overperform their PECOTA total, with well over half of the respondents picking them, while Cleveland were the second most popular choice to underperform. Despite that, Cleveland still edged the divisional picks battle, at 32-24, with another 12 picking them to take one of the wild card spots. There wasn’t as much strong feeling on the other three teams, but broadly speaking the consensus was that the Tigers would beat 79 wins (56.94%) while the Twins and White Sox would fail to reach 79 and 82 respectively (48.61% & 51.39%).

AL West

TeamPEC Feb 17PEC Mar 27FGDavBttPEssay
Astros888787919290
Rangers807980808784
Angels757681848377
Mariners848482878086
Athletics757680847588

Systems and humans alike spectacularly whiffed on the West last year, when the Angels and Mariners were the most popular picks for the division, so the Astros need to hope they’re not just as wrong this year. They’re the consensus first-place pick, and after that it’s pretty congested. Fangraphs has just two wins separating the other four teams, and while the systems tend towards Seattle in second place, the humans don’t. Texas in particular gets little respect from the projection systems relative to the writers. Even though opinion was split over PECOTA’s 88 win total for Houston – only 38.89% picked them to top that mark – the vast majority still picked them to win the division, although each of the other teams received at least one first place vote. Texas was the most popular wild card pick outside of the East, and that matched the 65.28% taking the over on 80 wins, with almost an identical amount taking the over on the Angels at 75. The listeners lean towards my take on the Mariners, with just over 70% indicating that 84 wins was too high, while there was a slight lean towards the under for the A’s.

NL East

TeamPEC Feb 17PEC Mar 27FGDavBttPEssay
Mets909090899191
Nationals878788888888
Marlins767580787375
Phillies656664647075
Braves686969666971

Last year 99% of listeners and all of our sets took the Nationals to win the division, so the three who took the Mets probably felt pretty smug. This year it’s a much closer race in the predictions and projections, which match up very closely across the board here – in fact, the writers predicting the Phillies to be a bit less terrible than the Braves is as close as we get to genuine disagreement on the Mets-Nationals-Marlins-Braves-Phillies order. The listeners couldn’t even muster a strong collective opinion on the over/unders: the closest they got was 55.56% taking the over on the Phillies at 65. There was a touch more disagreement in the playoff picks, however: Washington won the divisional battle 38-33, and there were a couple of bold Marlins picks, one for the division and one for the first wild card. Who knows; perhaps this year it’ll be their turn to be smug come October.

NL Central

TeamPEC Feb 17PEC Mar 27FGDavBttPEssay
Cubs92949510096102
Cardinals828185839589
Pirates838283858386
Reds747472677565
Brewers777870687374

As Sam lamented on the podcast last week, the NL is not the place to look for variance in predictions. The Central is barely more divided than the East: the Cubs are going to win, the Cardinals and Pirates will be close for second, and the Reds and the Brewers will battle not to be last. There is at least some disagreement on exactly how good (or bad) these teams will be within those brackets: PECOTA is oddly optimistic on the Brewers, for example, while Davenport sees both them and the Reds as barely better than the Phillies and Braves. The Crisafulli and Perry Cards predictions, and the Davenport/Armour/Levitt triple-digit Cubs, liven up what is otherwise a pretty consistent picture at the top. There’s also a gigantic 37-win spread between first place and last in the EW set as a result. The listeners certainly think PECOTA has gone reverse Royals with the Brewers, with over 83% taking the under on 77 wins, and almost as many (79.17%) doing the same for the Reds at 74. Much like last year, over 80% also took the over on PECOTA’s Pirates mark of 83, and nearly three-quarters backed the Cardinals to exceed 82. Despite the 92 win projection, more than half also took the Cubs over, with 62 picking them as division winners and six more giving them a wild card, which means that there were four respondents who don’t think Chicago will make the playoffs at all. Pittsburgh is the overwhelming favourite to be in the wild card game yet again, as 38 people picked them to take one of the spots.

NL West

TeamPEC Feb 17PEC Mar 27FGDavBttPEssay
Dodgers949492949594
Giants878588888990
Diamondbacks787881848383
Padres777674687579
Rockies747474677472

Hey, look! Another predictable division! I think we can all be reasonably confident that the West will finish Dodgers-Giants-Diamondbacks-Padres-Rockies, right? In other words, basically everyone thinks the division will just turn out the same as last year, with maybe the Rockies being slightly less terrible. The Giants are a little closer to the Dodgers this year than in 2015’s projections, and the Padres and Diamondbacks have swapped after Arizona took its shot at ‘winning the offseason’. The listeners certainly believe in Arizona more than PECOTA, with 73.61% taking the over on 78 wins, and almost 70% also thought the Dodgers should have fewer than 94, pushing them closer to the Giants. Similar numbers also thought the Padres (77) and the Rockies (74) would be more terrible than PECOTA expected. 57 still took LA to win the division, ahead of 14 for San Francisco, and one lone D-Backs optimist. The Giants were picked as the second wild card by a third of listeners, with 6 more giving them the first; it is an even year, after all.

Listeners vs PECOTA

The full, sortable summary of the listener responses to the PECOTA poll, and a Tableau visualisation of the over/unders can be seen below. Kansas City, Tampa Bay and Cleveland were comfortably the most popular picks to miss their projected totals, but the Rays were also one of four teams to be picked as the most likely on both sides. Atlanta and Minnesota both had just one voter on each side, while the White Sox were the only team to have multiple voters on both sides of the question. The size of the circles on Tableau graph indicate the relative total number of wins projected for each team, while the shade of the colours indicates whether listeners were more inclined to take the over (green) or the under (red). Hover over the circles to see the team and exactly how many of the 72 respondents agreed or took the over on PECOTA.

LeagueDivisionTeamPEC Feb 6% Over% Under% AgreeDivWC1WC2Over-PUnder-P
NLWestArizona7873.6112.513.8910420
NLEastAtlanta6827.7843.0629.1700011
ALEastBaltimore7270.8312.516.6700020
ALEastBoston8831.945018.062914601
ALCentralChicago9255.5623.6120.8382223
NLCentralChicago8234.7251.3913.89624220
NLCentralCincinnati745.5679.1715.2800000
ALCentralCleveland928.3383.338.333284021
NLWestColorado748.3373.6118.0600000
ALCentralDetroit7956.9419.4423.6163440
ALWestHouston8838.8930.5630.56621500
ALCentralKansas City7688.896.944.172479420
ALWestLos Angeles7566.6720.8312.531350
NLWestLos Angeles9420.8369.449.72571607
NLEastMiami7631.9443.062511001
NLCentralMilwaukee774.1783.3312.500003
ALCentralMinnesota7929.1748.6122.2221311
ALEastNew York8527.7836.1136.113316501
NLEastNew York9023.6152.7823.6147700
ALWestOakland7536.1141.6722.2220000
NLEastPhiladelphia6555.5611.1133.3300000
NLCentralPittsburgh8386.119.724.177251360
NLWestSan Diego7719.4465.2815.2800001
NLWestSan Francisco8745.8330.5623.611462400
ALWestSeattle8413.8970.8315.2811403
NLCentralSt. Louis8272.2213.8913.893101330
ALEastTampa Bay915.5690.284.176105129
ALWestTexas8065.2820.8313.89441310
ALEastToronto8659.7213.8926.393313700
NLEastWashington8747.222527.78389500

For a more qualitative perspective on how PECOTA is viewed by fans, I also asked all respondents to tell me which team they rooted for and whether the team’s fanbase thought that PECOTA hated their team. Only 12 fans responded yes, but they weren’t all Royals fans; in fact, only one was. The most well-represented team was Detroit, with three, and two Yankees fans also answered yes. Only two NL fans answered affirmatively, for Pittsburgh and San Francisco.

I also allowed fans to expand if they felt that the yes/no dichotomy was insufficient. Here are some of the responses:

“Hate is a strong word. We think they underrate us because our relief is so good.”

My team fan base doesn’t know what PECOTA is.”

Not as much as Ben and Sam.”

They probably hate their team enough on their own.”

Those four statements were from Angels, Braves, Pirates and Reds fans. I’ll let you guess which was which.

Composite & Adjusted Standings

As with last season’s projections and predictions, I’ve taken all of this year’s sets and averaged them into one composite win total, using the same method as before to scale the human predictions down to a set that would produce 2430 wins. I also included the standard deviation below as another way to see which teams had the narrowest and widest ranges. This year, I’ve created a new ‘Adjusted PECOTA’ prediction, by taking the original PECOTA projections and altering the win total by up to 10 wins in either direction, based on the percentage of voters taking the over or under. This was done simply by subtracting the under percentage from the over to get a net over/under figure, then dividing by 10, and modifying the win total by the resulting amount: for instance, the Royals had an 81.95% net over, so 8 wins have been added to their projection for a total of 84. These are included below in the ‘APEC’ column.

LeagueDivisionTeamPEC Mar 27FGDavBttPEssayStDevCompositeAPEC
ALWestAngels76818483773.1877979
ALWestAstros87879192902.0598889
ALWestAthletics76808475884.8838074
ALEastBlue Jays86848888871.4978690
NLEastBraves69696669711.66866
NLCentralBrewers78706873743.4417269
NLCentralCardinals81858395894.9648688
NLCentralCubs9495100961023.0729695
NLWestDiamondbacks78818483832.1358184
NLWestDodgers94929495940.989389
NLWestGiants85888889901.6738788
ALCentralIndians92868290913.7098784
ALWestMariners84828780862.5618378
NLEastMarlins75807873752.4827575
NLEastMets90908991910.7488987
NLEastNationals87888888880.48789
ALEastOrioles74797781812.6537778
NLWestPadres76746875793.6117472
NLEastPhillies66646470754.2146769
NLCentralPirates82838583861.478390
ALWestRangers79808087843.0338184
ALEastRays87818283852.1548382
ALEastRed Sox87888689841.728686
NLCentralReds74726775653.9297067
NLWestRockies74746774722.7137167
ALCentralRoyals75777090877.5217984
ALCentralTigers79818185802.048083
ALCentralTwins78788076832.3667877
ALCentralWhite Sox84808885832.6088380
ALEastYankees84828185841.478284

Now all that remains is to wait for the teams to play another 150 games or so, then we can get on with the important business of figuring out how wrong everyone was again. See you in October.

Next post:
Previous post:

Trackbacks/Pingbacks

  1.  2016 Midseason Projection Outliers | Banished to the Pen
  2.  Evaluating the 2016 Season Preview Predictions | Banished to the Pen
  3.  Analysing the 2017 Predictions & Projections | Banished to the Pen

Leave a Reply

Your email address will not be published.

This site uses Akismet to reduce spam. Learn how your comment data is processed.