Is Brian Dozier’s Power Repeatable? by Gerald Schifman February 15, 2017 When the offseason began, a Brian Dozier trade looked inevitable. The Twins’ second baseman was one of the majors’ most productive players last season, offering elite power, good baserunning, solid plate discipline and steady second-base defense. That all-around skill set would cost just $15 million over the next two years. So as spring training begins, why is Dozier still preparing to play for a rebuilding Twins club? Both a steep asking price and an abundance of good second basemen hindered a deal. But these reasons may not fully explain why Dozier is still bound for Fort Myers this month. After all, negotiations often begin with high asking prices before parties find a middle ground. Plus, even though the keystone is now a good offensive position, many teams would net an upgrade by acquiring Dozier. Perhaps there was more at play — namely, doubt in the minds of club executives that the pull– and fly-ball-prone Dozier could repeat his first-rate power-hitting. Consider what Bill James said about Dozier on the second-base edition of MLB Network’s “Top 10 Right Now” series: “You guys are too high on Dozier. A lot of those home runs are 360-footers that just skim over the wall. I don’t buy that.” As data from ESPN’s Home Run Tracker affirms, Dozier’s home runs are often unimpressive. Quality of Dozier’s 2016 Home Runs Tracker Stat Dozier’s Average Percentile, Hitters 10+ HR Percentile, Hitters 30+ HR True distance (feet) 396.7 33rd 10th Exit velocity (mph) 103.5 43rd 16th Spray angle 28° 20th 10th All percentiles are based on 2015–16 player-seasons.Spray angle is calculated as degrees away from the pull-side foul line for both RHB and LHB. By true distance, exit velocity, and spray angle, Dozier’s dingers often traveled shorter and at a lower velocity than his peers’ blasts, all while keeping closer to the pull-side foul line. When compared only to his power-hitting brethren — those with 30-homer campaigns — Dozier slips under the 17th percentile mark in every stat. Do the pedestrian home-run metrics hint at a coming power outage? Predicting Home-Run Rate To address that question about Dozier, we’ll first check whether the Home Run Tracker and other variables assist in projecting future home-run rate. I’ll fit a linear regression that predicts HR% (HR/PA) based on previous-year Home Run Tracker variables and batted-ball stats, as well as current-year Steamer HR% projections and contextual factors. Steamer’s projection backlog dates back to 2010, leaving 1,316 player-seasons that fit my minimum criteria. The data set was randomly split into two parts: a half for training and a half for testing. The model satisfies the assumptions of linear regression and performs relatively well. Its adjusted r-squared is 0.501 and out-of-sample RMSE is 1.06%; both figures represent improvements upon Steamer projections alone. HR% Regression Results Variable Estimate Std. Error P-value 95% Conf. Interval (Intercept) -20.726% 3.183% 0.000 (-26.975%, -14.476%) Projected HR%Y +0.676% 0.057% 0.000 (+0.565%, +0.788%) League HR%Y +1.356% 0.177% 0.000 (+1.010%, +1.703%) Avg. HR park factorY +0.019% 0.006% 0.001 (+0.008%, +0.030%) Avg. HR exit veloY-1 +0.144% 0.046% 0.002 (+0.054%, +0.234%) Airball%Y-1 +0.023% 0.008% 0.007 (+0.006%, +0.039%) Std. dev. of HR spray angleY-1 +0.036% 0.016% 0.022 (+0.005%, +0.067%) Pull%Y-1 +0.015% 0.009% 0.095 (-0.003%, +0.033%) Avg. HR spray angleY-1 -0.004% 0.013% 0.722 (-0.029%, +0.020%) Just enough%Y-1 +0.001% 0.003% 0.740 (-0.006%, +0.008%) Avg. HR true dist.Y-1 -0.002% 0.009% 0.806 (-0.019%, +0.015%) Y: Current yearY-1: Previous yearAirball%=1-GB%Park factors reflect the average un-halved bat-hand HR factor in the parks each hitter actually played at.Minimums: Players needed to have 200+ PA in the current year, and 200+ PA, 100+ BIP, and 8+ HR in the previous year to be included. The biggest drivers of HR% are a hitter’s Steamer projection and the league HR%. That latter factor controls for environment changes that raise the tide of all hitters. The third current-year variable, park factor, is also highly significant: the friendlier the home-run conditions, the better a batter’s eventual HR%. Also important are three of the “previous year” variables. All else equal, a player who homers at higher exit velocities will hit more dingers in the following year. Keeping the ball off the ground and launching homers to different angles around the diamond are also good signs for future power. Pull% isn’t significant at the conventional 0.05 level, but it’s influential: a Dozier-like tendency to hook the ball heightens a hitter’s future HR%. The last three variables — average spray angle, average true distance, and rate of “just enough” HRs — hold little predictive power, despite James’ concerns. Ruthian shots and dead-center homers are impressive, but it’s immaterial whether or not a batter regularly launches these blasts. Dozier is above average in some significant areas (like projected HR%, airball%, and pull%) but below average in others (like exit velocity and standard deviation of spray angle). Altogether, where does he come out in this model? Plugging in Minnesota’s most recent righty HR park factor and the 2016 league HR%, the model predicts Dozier’s HR% at… 4.62%. That’s higher than Steamer’s 4% prediction, so Dozier’s batting profile actually should aid his power effort. But 4.62% still would dip far below last year’s 6.08% rate. If Dozier stays healthy all through 2017, expect him to be a 30-homer guy rather than a 40-homer one. Dozier’s Performance on Balls in Play Now, Dozier’s slugging ability is wrapped up in more than homers — he also draws power from piling up doubles and triples. Could his performance on balls in play (BIP) also be headed for a downturn? Rather than plug the Home Run Tracker variables into another model, I’ll use two years of Statcast data to address whether Dozier enjoyed good fortune on BIP. For all righty batter-seasons in 2015–16, I sorted BIP into bins of 4º of launch angle and 2 mph of exit velocity (about one-sixth of a standard deviation in each measure). Then I found the average run value of the BIP in each bin. By mapping these average run values back to every batted ball, we can tally up to identify each hitter’s expected runs produced, based on contact quality. Between his actual singles, doubles, and triples last year, Dozier produced .313 runs/BIP. This is a fine measure that mirrored the league’s .314 mark across the two years. But his exit velocities and launch angles paint the picture of a lucky hitter, as his Statcast-based expectation was just .284. That 29-point differential is huge, and there’s little reason to think that Dozier can “beat” his Statcast profile like this again. For one thing, Dozier’s extreme pull tendencies largely limit him to one slice of the field and make him predictable for defensive positioning. For another, the correlation of hitters’ 2015 vs. 2016 Statcast-to-actual runs/BIP differentials (at 100 BIP minimums) was just 0.28. Thus, overperformance isn’t generally a reoccurring skill. To that point, Dozier’s runs/BIP in 2015 was a more modest .294, which was only a few points better than his Statcast expectation of .290. Perhaps a home-run loss will lead to a doubles gain, mitigating a decline on balls in play. But either way, Dozier can’t be expected to overshoot his “deserved” runs/BIP by 29 points again. As great as Dozier hit in 2016, teams’ hesitance to give up a bounty of prospects for his services is understandable. Dozier looks set for a drop-off, and the Twins may still struggle to justify a superstar-level return for their second baseman.