subvertadown
patreon
Personal update
🕑 Added 2020-10-26 13:19:08 +0000 UTCHi everyone, I thought it was a good time to tell you a little about what I have been doing. Normally my posts are all facts-only, and most of what I'm doing is hidden. I originally thought I might be writing more about ongoing work, but the fact is it sucks up a lot of time to prepare everything throughout each week. Here are a few reflections about the 2020 season and my activities since September.
First of all, thanks all for participating here on Patreon. As I've said before, it's not like I'm earning to make my whole living off this work, but it does mean a lot that you make the extra effort to express your appreciation. Some conversations got pretty toxic on Reddit last year, and actually there are still a lot of petty teenager comments coming through, time-to-time. Luckily, a lot of that has changed, but it is still somehow nice to have a little haven with purely appreciative followers, here.
About using Patreon itself: I was initially concerned with how to divide duties between here and Reddit. I think I've found an okay way to walk the fine line, between giving some extra value here while not withholding from Reddit. But there have been a few Patreons who left after I moved my "Tiered Flex But OK" off of Patreon. That's totally okay, as this space was intended really more for donations, as opposed to a quid pro quo exchange for service.
A lot of people are asking for the D/ST and kicker ROS charts, and so far they will remain only here. However, in week 11, I will post ROS charts for Reddit, because around that time it becomes essential for strategy.
About what I do: I really forgot just how much time gets taken up! There is so much manual work, which I would love to get automated. It begins Monday and Tuesday, when I begin collecting the results for each position-- I need to get the new weeks' data, both for making my accuracy analysis, as well as for using as input data. So I go to ESPN/NFL websites, find the position, copy the numbers and names, paste into Excel... check by eye that there's no error in the Excel formulas and conversion... because usually there is, and then copy and paste the numbers to a file with all the season tables. This means repeating it for 6 positions, and then I need to get all the other data for input: game scores, yards, sacks, etc. about 20 variables in total, each with its own table. Added to that, I need to do the same routine-- manually scraping websites for Vegas projections and weather forecast.
After collecting all that data, I then check the injury report so that I can make the most important tweaks to each team. This constant updating continues through the week: Check the updated Vegas betting lines, check the updated weather report, check updated injury reports....
When all the data is processed, I do the write-ups for my Tuesday and Wednesday posts-- 6 posts on Reddit (4 positional posts, 1 accuracy, 1 text-only post), plus 3 posts on Patreon. Every chart needs some manual typing in order to list the teams on bye. Then I copy paste to powerpoint, save the image as a jpg, transfer to reddit.
I nearly forgot, the accuracy comparisons.... since I decided to track so many sources for kicker and D/ST, I need to go to each of those websites and collect the rankings of all those sources. Usually there are errors in syntax etc. This season, every site uses something different for Washington, so I need to manually fix it for "Redskins" or "Team", otherwise the accuracy will not compute.
Let's not forget, trying to respond to comments (after skipping all the ones that have neglected to read my FAQ...)
Anyway, the main point is, I'm stretched pretty thin! There's a lot to maintain, and I haven't been able to put time into automating things or to writing updates for you all, here.
About the models: Overall, I'm pleased about both the performance of the models and their reception by the community. For examples, D/ST is doing just great, and people have been kinder to the Flex posts than I expected. But I do have a few frustrations: QB should have performed better in weeks 2&3, and it was just my own hubris to blindly trust my method and not calibrate inputs. Luckily, that period is over, especially as in-season data has overtaken, and now QB is looking good. Kicker-- although I'm still on top, there have been too many down weeks and some silly slip-ups. I feel I've proven that better models are possible, but I also still see ways to improve it further. But my main concern has been my game-score model, which I'll touch upon next time. I have recently improved it, but there's been additional interesting work going on there, and October has managed to fool me into thinking I can earn by betting. But more on that next time!
That's all for now, and good luck! I'll be back tomorrow with the ROS charts and tiebreakers...
Comments
Al Ha
It's all super appreciated! I wish I had the tech know-how to help you automate a lot of the raw data aggregation since it sounds like that would free up quite a bit of your time. Hopefully someone who can help will step up. I could see that being a kickstarter campaign - help fund a programmer to automate the data entry to keep the projections going! I would certainly contribute.
