Surrender Punts: A Look to the Future
Studying how DDSPF's 4th down decision making is slightly better in 2021
Congratulations everyone, you did it. You made it through 2020 Draft Day Sports Pro Football: 2016's reign over the ISFL Regular Season. And wow, what a time it's been. Obviously you all are here for surrender punts, but let's give a hand to how DDSPF16 managed to end the ASFC regular season. Truly marvelous. (More on this later!) Alas, our great Head Office has decided that it is time to move on from the possibly illegal? use of DDSPF16 to the newest generation of simulation football - DDSPF21! The folks at Wolverine Studios have had 5 years to update their programming and logic in the game, so let's see how it compares.
As a quick reminder, a surrender punt is a punt that is so cowardly that it is akin to surrendering the game. These typically happen late in games when the possessing team is at around the 30-35 yard line - what would be a 47-52 yard field goal - and has it on 4th down and less than 5 yards to go - prime go-for-it distance. You can review more about surrender punts and how they worked in DDSPF16 here on Medium.
Okay. Done reviewing? Awesome. Let's get started with the comparison.
Data Collection
In order to collect the data for this, I had to first get the DDSPF21 trial, import the ISFL players and teams, and then finally sim 10 seasons from the same identical save file. This took a surprisingly long time, but fortunately the trial period is for 1 week, granting me more than enough time to get everything set up and simmed.
The harder part was actually getting the index data. You see, the index generated by DDSPF21 doesn't spam out tons of HTML files for each game that happens - instead they use a handy little thing called LZ Compression to generate each game's box scores. The main box scores page will run a little script that looks for the correct game ID from the requested game in the compressed files, unpack it, and then display it. This means you can't use a simple web scraper like Beautiful Soup to get the data - instead you would have to run a browser on your own, have it load the page, and then scrape it. There's a way to do this in Python with Selenium, but it takes FOREVER, so at 3AM one morning I decided to try and figure out how to decompress the information from the files. Because there's a package for just about everything in Python, I was able to find a nice little package called lzstring that could do the hard work for me. It's significantly faster to scrape this way, taking about half a minute to scrape a season's worth of index. Using Selenium took OVER THIRTY MINUTES per season. Sitting there for 10 seasons worth was, well, I didn't because I let it run and then went and did something else.
Anyways, this is good because it means it's extremely simple to get the SurrenderBot upgraded to DDSPF21!
How Do Drives End?
In DDSPF16, nearly half (49.4%) of all drives ended with a punt, about 6% more than the NFL did (~43%).
Touchdowns and Field Goal Attempts followed thereafter, then interceptions, fumbles, turnovers on down, and safeties. This all makes sense. How does DDSPF21 do?
As we see here, a drastic decrease in the number of punts, down almost 8%. TDs, FGs, interceptions, and turnovers on down remain fairly consistent, but fumbles and safeties see a doubling in frequency. Either way, things seem to look basically the same between the two.
Decisions by Field Position
Probably the most dramatic chart from the original Surrender Punts article was the field position decision chart for 4th downs. There was a drastic change at the 40 and 30 yard lines, where punts basically sat between a 60% and 40% frequency between those two sticks. Past the 30 we saw zero punts and mostly field goals, as expected.
This is probably the biggest improvement I've seen for DDSPF21 - rather than the sharp, discontinuous cutoff at the 40 and 30, we see a much more continuous negative slope in punt frequency until we hit the 30 yard line.
This means instead of having a 40% chance of punting at the 31 yard line, there was a 20% chance - a big improvement! This is probably best displayed with a direct comparison:
The continuous form of this graph also looks much more similar to the NFL's curve shown in Jon Bois' video.
Do we see that the sim seems to have gotten very scared of going for it? Absolutely. I'm not sure why though.
Now, should anyone be punting at their opponent's 31 yard line? Heavens no. But look, we're going to take baby steps here. Also note, still no dramatic increase in tries at the 1 yard line.
The Dead Zone
In the first article, I dubbed DDSPF16's lines in the sand between the 30 and 40 as "The Dead Zone". As we saw above, the Dead Zone still has some spooky action going on, though it's not as drastic as it was in the past.
The first big improvement in Dead Zone logic comes with the distance needed for a first down. In DDSPF16, we saw that punts took up a relatively constant proportion of decisions made up until 4th and 10 - at which punts took up a significant majority of the plays.
Compare this to DDSPF21:
In the new version of the game, tries and field goals make up a much larger amount of the proportion, even going past 4th and 10. On 4th and 1, punts are almost an impossibility between these two yard lines! Some of the most dramatic surrender punts in ISFL history occurred on 4th and less than 3, and so seeing this data is extremely encouraging.
Now, what effect does the point differential play to this? In DDSPF16, we saw that there was some questionable logic in the 4th Quarter:
Now, punting in this zone while tied doesn't make a ton of sense - and neither does going for it! A 50ish yard field goal has a very good chance of converting (more on this later), and so giving your opponent the chance to take the lead instead makes little sense. As we've seen before, the sim also likes to punt when down by other scores as well... which, why?
Here is DDSPF21:
By far the biggest difference here is that the sim is much much more in tuned with kicking the field goal here. When down by less than 3, kicking the field goal is DDSPF21's favorite option. However, when down by 3, it reverts back to a heavy-punt[/color] decision making strategy. Why? Who knows. Down 4 it goes back to kicking the field goal (????) before favoring tries for pretty much any margin afterwards.
I think overall, we can say the logic in the Dead Zone is significantly better in DDSPF21, but there are still things to be worked out.
Going For It
The success rate of going for it on 4th down does not seem to have changed much between the two games. Here is DDSPF16:
And here is DDSPF21:
The main difference here is that tries become much less common when it takes more than a yard to get the first down. This makes sense, and is a good improvement! There's no reason a 4th and 10 should have as many attempts as a 4th and 5. The drastic drop past 4th and 4 still exists, though it is much more noticeable in the DDSPF21 plot.
Overall, going for it on 4th down seems to have improved where it matters most - distance. Nothing much to add here.
Field Goals
Look, kickers in this league are going to care a lot more about whether or not they can make field goals than they care about these punts. So, let's take a look at field goal accuracy and frequency between the two sims.
Here is DDSPF16:
The Dead Zone strikes again - the number of field goals decreases dramatically at the 30-ish yard line and continues as such until we get to 55 yard field goals. The overall rate of conversion seems pretty reasonable though, with kickers hitting around 80% between 50 and 55 before dipping towards 60% past 55. (Anything past 60 on the top chart can't really be taken seriously due to sample size - see the second plot!)
How does DDSPF21 stack up?
The Dead Zone's effects are again less drastic in DDSPF21, with a nice and more gradual downward slope past 50 yard field goals. It does seem like kickers are worse in the newer sim: between 50-55 yards they drop to about 70%, and even from 40-50 yards they seem to miss more often. Once again, don't take anything on the top chart past 60 yards very seriously, as sample size strikes.
In both sims, the longest possible field goal attempt appears to be 65 yards.
Overall, field goals make slightly more sense, but expect to miss more often from distance.
Punt Results
In an article about Surrender Punts, you should expect to see some talk about punting results, right? Well, I have good news here too! You may recall that arguably the most horrendous and offending part of the high rate of surrender punts was that just about every single punt snapped inside the 40 ended in a touchback:
Why program in a 50% punt rate when it always results in a touchback? It was thoroughly mind boggling to me. Well, it seems like my cries were heard for DDSPF21:
The COFFIN CORNER exists now! While the surrender formula (see the original article) does not account for the result of the punt, this can possibly help those who do a surrender punt to feel a little bit better about themselves. About 20% of punts taken inside the 35 end up as inside the 20, and this rises up to nearly 40% as we approach the 40. That's over a 35% jump!!! Touchbacks obviously remain the most common occurrence, but the fact that punters can now use accuracy to pin opponents deep in their own territory is a huge improvement.
Overall, I can say that the logic within punts themselves is vastly improved, with coffin corners now being an option.
Surrender Punts!
As of December 1, 2020 - 25 seasons - we had 7 punts that were worse than the NFL's most cowardly punts. How did we do in my 10 simmed seasons?
Time: 4th Quarter, 2:49 remaining
Score: 24 - 28
Punting Team:
Down and Distance: 4th and 2
Field Position: 32
Result: Punt by McDairmid, M. of 30 yards. No return. (Downed at the 5)
Final Score: 31 - 28
Thanks to the new c o f f i n c o r n e r mechanic, this objectively horrible surrender punt still gets to call itself a success! McDairmid's punt landed at the Wraith's 5 yard line, where they proceeded to run three times, end up at 4th and 11, and punt it right back to the cardiac kitties. Now, they start way back at their own 43 now, but they manage to march down the whole distance to score a touchdown with 38 seconds left to go! What a success for the sim! But wait, there's more!
What on God's green Earth is THIS?! Draft Day Sports. Calling timeouts at reasonable times. IMPOSSIBLE! Now it's not all sunshine and rainbows - a few plays later the sim takes 40 seconds to do a 1 yard rush, and then another 25 seconds for another 1 yard rush.
Baby steps I guess.
Time: 4th Quarter, 4:35 remaining
Score: 9 - 23
Punting Team:
Down and Distance: 4th and 3
Field Position: 32
Result: Punt by QB ®, P. of 32 yards. Touchback.
Final Score: 16 - 25
Yet another surrender punt where it kind of works out for the punting team. But only kind of - when you're down by 14 you really shouldn't be punting the ball back. The formula punishes the Copperheads adequately for this, and even with the touchback making it a net punt of 12 yards, the Copperhead defense stands strong and forces the Wraiths to punt after a three and out - before immediately going down the field and scoring a touchdown in two plays. Their onside kick after failed, but they manged to get the ball back after another 3 and out - but this time, pinned in their own 10. Desperate and with 23 seconds to go on 2nd and 11, they look to complete the big play when Easton Cole is dragged down for a 10 yard loss and a safety. You can only imagine what would have happened if they didn't lose 90 seconds giving up possession with this surrender punt.
Time: 4th Quarter, 6:08 remaining
Score: 17 - 24
Punting Team:
Down and Distance: 4th and 2
Field Position: 30
Result: Punt by McDairmid ©, M. of 26 yards. No return. (Downed at the 4)
Final Score: 24- 27
Now this is absolutely wild stuff here. On the onset, you probably saw that there were 6 minutes remaining - how on Earth is this worse than anything we've seen in TWENTY FIVE SEASONS of DDSPF16? Well my friend, this was punted at the opponent's 30 yard line. Down 7. On 4th and 2. With 6 minutes to go. Truly pathetic.
But okay, there's still lots of time, and hey, McDairmid pinned them in their own 5! On 1st and 10, the Yeti rush for a gain of 2. But then, Wolfie McDummy takes a sack and barely sneaks out of his own endzone! Now at their own 1 yard line, the Yeti try to sneak out of the endzone when they're tackled for a safety! Great success! The following free kick is only kicked 35 yards (free kicks don't seem to go very far in this sim), and the heart attack inducing housecats march down the field, burn 3 minutes of clock, and score what is surely the game winning touchdown! The 2 point conversion failed, but they're up by 1 with 1:28 to go. The Yeti get the ball back and after a 26 yard return by Louisiana Purchase, have to start at their own 26. With 56 seconds left, the Yeti are stuck with a 3rd and 4 when a Zobot gets dinged with an Unnecessary Roughness penalty! This, on top of the 6 yard completion, gets the Yeti across the 50 with 41 seconds to go. McDummy hits Robertkowski for a quick 9 yard gain before spiking the ball with 22 seconds left. Then, 19 seconds mysteriously vanish off the clock (spikes seem to be weird with time in this sim too) before Silver Banana hits the 53 yard field goal to take the lead! Tychondrious Hood can only return the ensuing kickoff 37 yards, and the game is over.
Truly a wild ending. The coffin corner uber surrender punt almost worked out for them if not for a Zobot being stupid and risking a concussion or something.
Also, can we just note that Matthew McDairmid, the Sabercats kicker, is the strongest kicker in the game. Of the 15 kicks attempted from over 60 yards, McDairmid attempted 10 of them (including all 6 from 64 and 65) and made 6 of them. In DDSPF21, McDairmid went 62/87 from 47+. WHY ARE THE KITTIES TWO OF THE TOP THREE HERE?
Now, that would have gone out as the most cowardly surrender punt in the history of surrender punts... but boy oh boy did DDSPF16 have one last wild surprise for us.
Time: Overtime, 8:40 remaining
Score: 20 - 20
Punting Team:
Down and Distance: 4th and 2
Field Position: 32
Result: Punt by Jay-Jaymison, J. of 32 yards. Touchback.
Final Score: 26 - 20
The Outlaws were on the outside looking in when it came to the ASFC playoffs - sitting at 8-7 and going to face the 9-6 Second Line. While they were mathematically eliminated from the playoffs, they had the chance to play spoiler for the Second Line, and possibly help create the greatest standings finish ever seen. After forcing overtime with 54 seconds to go, the Outlaws overcame a classic Jay Cue interception to get the ball back at their own 37. Any score would win the game. I repeat, ANY SCORE WOULD WIN THE GAME. And so, the Outlaws marched. They marched right into field goal range. And on 4th and 2 at the New Orleans 32 yard line, in J.J. Jay Jay-Jaymison's field goal range, decide to have him punt. J5 was 9/9 from 40-49 yards, and had a long of 48 on the season. This was a 49 yard field goal attempt. When any score could have won the game. But alas, they go for the 12 yard net play and let New Orleans have the ball instead. They are quickly saved by a bot offensive lineman committing an Unsportsmanlike Conduct penalty, forcing the Second Line to punt. Granted a second chance by some deity who wants to see chaos, the Outlaws march right back down the field thanks to a 45 yard throw and catch, scoring the winning touchdown with 4:21 to go.
Not only was this the absolute worst surrender punt I have ever ever seen, it WORKED. Which is just wrong. But I think the sim gods had the bigger picture in mind - this result forced a 5-way tie for first place in the ASFC, unleashing the ultimate chaos and allowing New York, New Orleans, and somehow Orange County to advance to the playoffs. Had Arizona not won this game, it would be a 3 way tie between New York, Orange County, and Austin for 2nd, and after New York advanced on head to head, Orange County would have not have had the tiebreaker over Austin thanks to conference record. With Arizona (and New Orleans) in the mix, the Otters' sweep of the Outlaws allowed them to jump the Copperheads, who split their games with the Outlaws and Otters.
So while surrender punts should always fail, the sim gods gave us one final gift (or curse if you are a Copperhead) by allowing this to work out.
Hey, if you want to see the scraping code I used, hit me up. The Jupyter notebook is a bit of a mess, but if there's interest I can clean it up. The DDSPF21 Indexes for this project are hosted on my personal webpage, follow one of the links above to get there.
You can also read this article on Medium! https://medium.com/the-specialist/surren...d5e780b8ae
Also this is over 3,000 words and somehow I missed double media pay. Sad days.