This Is Your Pilot Speaking: Flying into the Future and Predicting this Week’s MLTP Results
by Ball Simmons
This Is Your Pilot Speaking: Flying into the Future and Predicting this Week’s MLTP Results
After its introduction last season, Pilot quickly became one of the most interesting maps, due to a mix of complexity and simplicity. A casual glance at a preview of the map would lead an observer to assume straightforward gameplay, but the limited elements lend themselves to a game of complexity. The primary of these are the bombs in the middle of the map, which tempt risky balls to ricochet away into their opponents. It seems at times that most important factor in a Pilot match is making the correct decision.
I’m going to forget the nuanced decisions of the map to focus on what’s tangible: stats. Measurements like GASP and NISH value players who are productive, lots of returns or hold are a good thing, right? When slide inevitably puts up over seventy tags this week, should we begin to hand him the award for defensive player of the season? Fortunately, slide is also on a good team, and as MLTP success relies heavily on a team, the Base Gods are a very productive team as well by leading the league in returns and captures.
But let’s ignore what teams have already done over the entire course of the season, and focus on what they have yet to do. Let’s predict the future. In the past I modelled team success based on the totals for each stat, which worked pretty well, but was flawed because of teams like LagProne which have barely any hold, but lots of returns and not that much hold against. If a team prevents its enemy from having the flag a lot, it matters less that a team doesn't have its opponent’s flag out, especially if they capitalize on scoring opportunities.
Proportions, then, were the best way to combat this. Simply, the total of a teams production divided by their plus the opponent’s production.(team’s total)/(team’s total + opponent’s total), gave the proportion of that stat total earned by the team. Each week two maps are played, so Pilot was chosen because it is again played Week 7, so we can use last week's stats to predict next week’s results. Two models were created using backwards elimination to calculate plus/minus (how many net caps more (or less) a team scores compared to their opponent) based on the proportions for each half of Pilot played the previous week. Data from the previous week was used to predict each team’s stats for week 7, and the proportions for the upcoming meetups were fed into the models to create the predictions.
The first model is from an earlier step in the backwards elimination, which means not all variables have the strongest correlation, but the model had the lowest error. The second model comes from the end so all variables are strongly correlated to plus/minus (>0.05), but has fewer independent variables and slightly higher error.
The strongest influences for both models are grabs and drops, which is very similar to captures (a variable manually removed from the data). The negative strength of drops also ties into pops, because every time a player pops they are sidelined for 3 seconds. Should a player pop 20 times, they would be inactive for 10% of the game. Thus living proportionally more than the opponent is a huge factor in scoring having a positive plus/minus, due to the overtime player advantage. The first model also includes returns, prevent, and hold against. The strangest part of the model is the negative weight for defensive statistics: prevent and returns. Prevent is very weak anyway, but is likely affected by LagProne’s tied half despite having only 20.8% of the prevent. Returns however, is different, and probably stems from the fact that a team with a large proportion of the returns will have a small proportion of the grabs, and grabs are a huge positive (if you’re not dropping a lot too). Additionally, Pilot is unique in that many drops are caused by flag carriers spiking due to mid bombs which makes a standard return less important.
With this in mind, what predictions will come true?
Origin Ducks are comfortable in the rankings right now, but lost to the dominant Merballs on Pilot. Roll Models on the other hand are making up ground with victories like the one over Holdin’ Gate Warriors. These teams end up neck in neck in the predictions however. Neither team is excessively grabby, but the Roll Models have a slight edge from not popping or dropping as much. What gives the Origin Ducks the slightest of edges in Model One, is their lower share of returns which matches their higher grab proportion. The Ducks are also predicted to win powerups, but whichever team can control them better might have the edge to send it one way or the other.
Merballs should wrap up another victory against the Holdin’ Gate Warriors who are their biggest threats to the Western crown. While the Merballs are already exceptional at staying alive more than their opponents, Holdin’ Gate struggled and as a result are predicted to have twice as many non-drop-pops.
Angry Balls Redux against Tears should be another close match, with the Redux coming off of a dominant victory over Texas Hold ‘Em and Tears sliding past BMX2. Interestingly, the lowest pops Tears had on a half last week (55) is equal to the Angry Balls’ max. However Tears’ powerup strength saves them in the prediction. Angry Balls will likely dictate the flow of this game but Tears should use powerups to get the caps needed to keep it close.
The last western conference matchup predicts a low ranked BMX (who played Tears close) to take a game off of middling Texas Hold ‘Em. Both teams are nearly identical in grabs and drops, but Texas Hold ‘Em was on the upper end of pops. Additionally BMX should rebound with better pup control and take care of the game.
The most exciting prediction of the night is that the first model expects LagProne to bring the Base Gods down to demigods with their first defeat. LagProne is very conservative with grabs, they averaged 30 per half, and and while they popped a bit more than that, they were still on the lower side of the league. Base Gods on the other hand are volatile and explosive, which is what makes them exciting. Though they aren't far above league average in grabs, drops, or pops. The second model predicts a Base Gods victory, as the first model gives them a hard time for inflated defensive statistics. The stingy LagProne offense won’t feed the slide return machine so it would be very surprising to see Base Gods’ offenders past 3. Both teams are strong with powerups but LagProne will need to gain control to make the most of their limited chances.
The first model likes the KGBallers a whole lot more than the second, but KGB had a close game against LagProne whereas Adrenaline Spikes were blown out a bit by Base Gods. AS shattered every other team by averaging 74.5 pops which does not bode well for their chances. They did this by grabbing a lot, so if they manage to convert they could make it close, otherwise they’ll definitely be putting this season to rest. Xile doesn’t help this problem much either, he plays both positions and would be more aggressive on offense than a standard defender.
The new and improved Dyson Majors are back on track but will have their hands full with pimped out Boostin Dynamo. jjpoole was able to fleece his former teammates and upgrade his positions of need. This prediction is likely to be inaccurate, especially due to the fact that both teams were very close statistically.
Ghostboosters and ALL CAPS both lost last week on Pilot but Ghostboosters managed one good half against Dyson before falling into shambles for the second. Neither team stands out much except GB’s powerup control. ALL CAPS pop a little less but still drop more. Close game but this homecoming game should be exciting nevertheless.
After its introduction last season, Pilot quickly became one of the most interesting maps, due to a mix of complexity and simplicity. A casual glance at a preview of the map would lead an observer to assume straightforward gameplay, but the limited elements lend themselves to a game of complexity. The primary of these are the bombs in the middle of the map, which tempt risky balls to ricochet away into their opponents. It seems at times that most important factor in a Pilot match is making the correct decision.
I’m going to forget the nuanced decisions of the map to focus on what’s tangible: stats. Measurements like GASP and NISH value players who are productive, lots of returns or hold are a good thing, right? When slide inevitably puts up over seventy tags this week, should we begin to hand him the award for defensive player of the season? Fortunately, slide is also on a good team, and as MLTP success relies heavily on a team, the Base Gods are a very productive team as well by leading the league in returns and captures.
But let’s ignore what teams have already done over the entire course of the season, and focus on what they have yet to do. Let’s predict the future. In the past I modelled team success based on the totals for each stat, which worked pretty well, but was flawed because of teams like LagProne which have barely any hold, but lots of returns and not that much hold against. If a team prevents its enemy from having the flag a lot, it matters less that a team doesn't have its opponent’s flag out, especially if they capitalize on scoring opportunities.
Proportions, then, were the best way to combat this. Simply, the total of a teams production divided by their plus the opponent’s production.(team’s total)/(team’s total + opponent’s total), gave the proportion of that stat total earned by the team. Each week two maps are played, so Pilot was chosen because it is again played Week 7, so we can use last week's stats to predict next week’s results. Two models were created using backwards elimination to calculate plus/minus (how many net caps more (or less) a team scores compared to their opponent) based on the proportions for each half of Pilot played the previous week. Data from the previous week was used to predict each team’s stats for week 7, and the proportions for the upcoming meetups were fed into the models to create the predictions.
Prediction % InfographicThe first model is from an earlier step in the backwards elimination, which means not all variables have the strongest correlation, but the model had the lowest error. The second model comes from the end so all variables are strongly correlated to plus/minus (>0.05), but has fewer independent variables and slightly higher error.
The strongest influences for both models are grabs and drops, which is very similar to captures (a variable manually removed from the data). The negative strength of drops also ties into pops, because every time a player pops they are sidelined for 3 seconds. Should a player pop 20 times, they would be inactive for 10% of the game. Thus living proportionally more than the opponent is a huge factor in scoring having a positive plus/minus, due to the overtime player advantage. The first model also includes returns, prevent, and hold against. The strangest part of the model is the negative weight for defensive statistics: prevent and returns. Prevent is very weak anyway, but is likely affected by LagProne’s tied half despite having only 20.8% of the prevent. Returns however, is different, and probably stems from the fact that a team with a large proportion of the returns will have a small proportion of the grabs, and grabs are a huge positive (if you’re not dropping a lot too). Additionally, Pilot is unique in that many drops are caused by flag carriers spiking due to mid bombs which makes a standard return less important.
Pilot Prediction Model ResultsWith this in mind, what predictions will come true?
Origin Ducks are comfortable in the rankings right now, but lost to the dominant Merballs on Pilot. Roll Models on the other hand are making up ground with victories like the one over Holdin’ Gate Warriors. These teams end up neck in neck in the predictions however. Neither team is excessively grabby, but the Roll Models have a slight edge from not popping or dropping as much. What gives the Origin Ducks the slightest of edges in Model One, is their lower share of returns which matches their higher grab proportion. The Ducks are also predicted to win powerups, but whichever team can control them better might have the edge to send it one way or the other.
Merballs should wrap up another victory against the Holdin’ Gate Warriors who are their biggest threats to the Western crown. While the Merballs are already exceptional at staying alive more than their opponents, Holdin’ Gate struggled and as a result are predicted to have twice as many non-drop-pops.
Angry Balls Redux against Tears should be another close match, with the Redux coming off of a dominant victory over Texas Hold ‘Em and Tears sliding past BMX2. Interestingly, the lowest pops Tears had on a half last week (55) is equal to the Angry Balls’ max. However Tears’ powerup strength saves them in the prediction. Angry Balls will likely dictate the flow of this game but Tears should use powerups to get the caps needed to keep it close.
The last western conference matchup predicts a low ranked BMX (who played Tears close) to take a game off of middling Texas Hold ‘Em. Both teams are nearly identical in grabs and drops, but Texas Hold ‘Em was on the upper end of pops. Additionally BMX should rebound with better pup control and take care of the game.
The most exciting prediction of the night is that the first model expects LagProne to bring the Base Gods down to demigods with their first defeat. LagProne is very conservative with grabs, they averaged 30 per half, and and while they popped a bit more than that, they were still on the lower side of the league. Base Gods on the other hand are volatile and explosive, which is what makes them exciting. Though they aren't far above league average in grabs, drops, or pops. The second model predicts a Base Gods victory, as the first model gives them a hard time for inflated defensive statistics. The stingy LagProne offense won’t feed the slide return machine so it would be very surprising to see Base Gods’ offenders past 3. Both teams are strong with powerups but LagProne will need to gain control to make the most of their limited chances.
The first model likes the KGBallers a whole lot more than the second, but KGB had a close game against LagProne whereas Adrenaline Spikes were blown out a bit by Base Gods. AS shattered every other team by averaging 74.5 pops which does not bode well for their chances. They did this by grabbing a lot, so if they manage to convert they could make it close, otherwise they’ll definitely be putting this season to rest. Xile doesn’t help this problem much either, he plays both positions and would be more aggressive on offense than a standard defender.
The new and improved Dyson Majors are back on track but will have their hands full with pimped out Boostin Dynamo. jjpoole was able to fleece his former teammates and upgrade his positions of need. This prediction is likely to be inaccurate, especially due to the fact that both teams were very close statistically.
Ghostboosters and ALL CAPS both lost last week on Pilot but Ghostboosters managed one good half against Dyson before falling into shambles for the second. Neither team stands out much except GB’s powerup control. ALL CAPS pop a little less but still drop more. Close game but this homecoming game should be exciting nevertheless.