Ben Griffis
@begriffis.bsky.social
Analyst at Kolding IF 🇩🇰 | ex. Loudoun United 🇺🇸 | Views my own
I read every state's laws in 2019, so it's very possible more states have changed their laws/removed 3-tier laws. When I collected this data, no states that had repealed the 3-tier system had then re-enacted it, I'd assume that's still the case
October 6, 2025 at 6:00 PM
I read every state's laws in 2019, so it's very possible more states have changed their laws/removed 3-tier laws. When I collected this data, no states that had repealed the 3-tier system had then re-enacted it, I'd assume that's still the case
Yeah 100%. In grad school I was working on research around when states do away with the 3-tier law, and it improved production & increased the number of new breweries. Which is obvious, but crazy that states continue to limit their own economy
(unpublished & unfinished, but key results in image)
(unpublished & unfinished, but key results in image)
October 6, 2025 at 3:17 PM
Yeah 100%. In grad school I was working on research around when states do away with the 3-tier law, and it improved production & increased the number of new breweries. Which is obvious, but crazy that states continue to limit their own economy
(unpublished & unfinished, but key results in image)
(unpublished & unfinished, but key results in image)
This is why I think Import beer continues to increase despite craft beer sales and consumption falling
Virtually all the imports are "regular" beers, not hopped to the max, not sweetened, not including random experimental foods added in the brewing process
Virtually all the imports are "regular" beers, not hopped to the max, not sweetened, not including random experimental foods added in the brewing process
October 6, 2025 at 9:33 AM
This is why I think Import beer continues to increase despite craft beer sales and consumption falling
Virtually all the imports are "regular" beers, not hopped to the max, not sweetened, not including random experimental foods added in the brewing process
Virtually all the imports are "regular" beers, not hopped to the max, not sweetened, not including random experimental foods added in the brewing process
As a huge fan and proponent of craft beer, I feel like a few things are going on:
1) obv, less beer consumption to begin with
2) Too many microbreweries went extreme with the additives and/or over-hopping
3) At some point, experiments replaced quality
www.nytimes.com/2025/10/05/b...
1) obv, less beer consumption to begin with
2) Too many microbreweries went extreme with the additives and/or over-hopping
3) At some point, experiments replaced quality
www.nytimes.com/2025/10/05/b...
October 6, 2025 at 9:33 AM
As a huge fan and proponent of craft beer, I feel like a few things are going on:
1) obv, less beer consumption to begin with
2) Too many microbreweries went extreme with the additives and/or over-hopping
3) At some point, experiments replaced quality
www.nytimes.com/2025/10/05/b...
1) obv, less beer consumption to begin with
2) Too many microbreweries went extreme with the additives and/or over-hopping
3) At some point, experiments replaced quality
www.nytimes.com/2025/10/05/b...
Then, weight your metrics! I personally like mine to all add up to 1, but it's not necessary. It's all the same calculation as the role ranks built into the app - normalized weighted z-scores in each metric. So the top player is always 100, the others based off their score
September 9, 2025 at 2:38 PM
Then, weight your metrics! I personally like mine to all add up to 1, but it's not necessary. It's all the same calculation as the role ranks built into the app - normalized weighted z-scores in each metric. So the top player is always 100, the others based off their score
Finally, you can also create your own roles with metric ratings!
Choose what position(s) to include, then the metrics
Choose what position(s) to include, then the metrics
September 9, 2025 at 2:38 PM
Finally, you can also create your own roles with metric ratings!
Choose what position(s) to include, then the metrics
Choose what position(s) to include, then the metrics
There's some new features in the player search filters, again choosing between league/region/continent and then choosing seasons/tiers as needed. I tend to just use the League function but it could help to just say "Europe" and all 1st tier leagues for ease
September 9, 2025 at 2:38 PM
There's some new features in the player search filters, again choosing between league/region/continent and then choosing seasons/tiers as needed. I tend to just use the League function but it could help to just say "Europe" and all 1st tier leagues for ease
There's no guarantee it'll be perfect, by the way. I will still be tweaking with the methods to try to make it better. But it could help you find some interesting players for now
September 9, 2025 at 2:38 PM
There's no guarantee it'll be perfect, by the way. I will still be tweaking with the methods to try to make it better. But it could help you find some interesting players for now
The similarity is built on IDs. The league-season that the focal player played in has to be loaded... and to find the player's ID, head to the Player List tab, hover the table & click the magnifying glass to search, and copy the player's ID into the similarity search tab
September 9, 2025 at 2:38 PM
The similarity is built on IDs. The league-season that the focal player played in has to be loaded... and to find the player's ID, head to the Player List tab, hover the table & click the magnifying glass to search, and copy the player's ID into the similarity search tab
Another new feature is searching for similar players using Cosine similarity
Choose from options like league/region/continent, positions to include, and then either a custom set of metrics (I recommend this) or a pre-made set
Choose from options like league/region/continent, positions to include, and then either a custom set of metrics (I recommend this) or a pre-made set
September 9, 2025 at 2:38 PM
Another new feature is searching for similar players using Cosine similarity
Choose from options like league/region/continent, positions to include, and then either a custom set of metrics (I recommend this) or a pre-made set
Choose from options like league/region/continent, positions to include, and then either a custom set of metrics (I recommend this) or a pre-made set
... but now you CAN change what positions are included in the sample!
And as a bonus, if you want to compare a striker to wingers, you don't need to include both STs & Ws. The app will automatically put the player in the sample if they play another position than what's selected
And as a bonus, if you want to compare a striker to wingers, you don't need to include both STs & Ws. The app will automatically put the player in the sample if they play another position than what's selected
September 9, 2025 at 2:38 PM
... but now you CAN change what positions are included in the sample!
And as a bonus, if you want to compare a striker to wingers, you don't need to include both STs & Ws. The app will automatically put the player in the sample if they play another position than what's selected
And as a bonus, if you want to compare a striker to wingers, you don't need to include both STs & Ws. The app will automatically put the player in the sample if they play another position than what's selected
For pre-made radars, you'll now be able to select templates to use. These only change the metrics, not the positions included in the sample
September 9, 2025 at 2:38 PM
For pre-made radars, you'll now be able to select templates to use. These only change the metrics, not the positions included in the sample
On the Player Radar Generation tab, you'll notice that the Custom radar option is back. And better than ever! There used to be issues when changing variables after the initial custom selection... no more! Create your own custom radars with ease
September 9, 2025 at 2:38 PM
On the Player Radar Generation tab, you'll notice that the Custom radar option is back. And better than ever! There used to be issues when changing variables after the initial custom selection... no more! Create your own custom radars with ease
The women's app, however, has some key updates and added functionality
1st, on sidebar you'll see a maximum minutes option... this impacts every tab. This could help you find & rank bit-part players, say between 400 & 800 minutes only. Defaults to 5000, leave there if not needed
1st, on sidebar you'll see a maximum minutes option... this impacts every tab. This could help you find & rank bit-part players, say between 400 & 800 minutes only. Defaults to 5000, leave there if not needed
September 9, 2025 at 2:38 PM
The women's app, however, has some key updates and added functionality
1st, on sidebar you'll see a maximum minutes option... this impacts every tab. This could help you find & rank bit-part players, say between 400 & 800 minutes only. Defaults to 5000, leave there if not needed
1st, on sidebar you'll see a maximum minutes option... this impacts every tab. This could help you find & rank bit-part players, say between 400 & 800 minutes only. Defaults to 5000, leave there if not needed
The men's app is the exact same, but without a gender selection dropdown (since now they are separate apps/sections)
September 9, 2025 at 2:38 PM
The men's app is the exact same, but without a gender selection dropdown (since now they are separate apps/sections)
This is the landing page. On mobile, you'll probably need to click the << in the top left to see the sidebar
Choose which gender you want to analyze
Choose which gender you want to analyze
September 9, 2025 at 2:38 PM
This is the landing page. On mobile, you'll probably need to click the << in the top left to see the sidebar
Choose which gender you want to analyze
Choose which gender you want to analyze
In the 2nd half VT made adjustments. Virginia's strikers were caught a little lazy, but the GK was a bit more direct playing the ball up to the forwards. Some good play & poor UVA defending leads to a great opportunity
September 8, 2025 at 1:15 AM
In the 2nd half VT made adjustments. Virginia's strikers were caught a little lazy, but the GK was a bit more direct playing the ball up to the forwards. Some good play & poor UVA defending leads to a great opportunity
Thought it was really interesting how UVA sent both strikers to press the GK on goal kicks against Virginia Tech
UVA's strikers effectively blocked both CBs as passing options. VT's GK played a vertical ball to a CM who then passed to the RCB. But UVA kept the press up very well
UVA's strikers effectively blocked both CBs as passing options. VT's GK played a vertical ball to a CM who then passed to the RCB. But UVA kept the press up very well
September 8, 2025 at 1:15 AM
Thought it was really interesting how UVA sent both strikers to press the GK on goal kicks against Virginia Tech
UVA's strikers effectively blocked both CBs as passing options. VT's GK played a vertical ball to a CM who then passed to the RCB. But UVA kept the press up very well
UVA's strikers effectively blocked both CBs as passing options. VT's GK played a vertical ball to a CM who then passed to the RCB. But UVA kept the press up very well
A sample:
Here's the Ekstraklasa's next round of fixtures predicted with current team ratings
3 games are predicted to be extremely close while Korona Kielce & Lech Poznań are predicted to be heavy favorites
Here's the Ekstraklasa's next round of fixtures predicted with current team ratings
3 games are predicted to be extremely close while Korona Kielce & Lech Poznań are predicted to be heavy favorites
September 6, 2025 at 1:24 PM
A sample:
Here's the Ekstraklasa's next round of fixtures predicted with current team ratings
3 games are predicted to be extremely close while Korona Kielce & Lech Poznań are predicted to be heavy favorites
Here's the Ekstraklasa's next round of fixtures predicted with current team ratings
3 games are predicted to be extremely close while Korona Kielce & Lech Poznań are predicted to be heavy favorites
The paper published the RPS of 9 competition entries. The competition did have a separate training dataset so not perfectly comparable, but I pulled the testing set (206 matches in April 2017) and ran my model on it
My model performed better than all of them, weirdly enough
My model performed better than all of them, weirdly enough
September 6, 2025 at 1:24 PM
The paper published the RPS of 9 competition entries. The competition did have a separate training dataset so not perfectly comparable, but I pulled the testing set (206 matches in April 2017) and ran my model on it
My model performed better than all of them, weirdly enough
My model performed better than all of them, weirdly enough
Here are the top 10 biggest upsets in T5 last season
RPS works off of pre-match probabilities of a home win, draw, & away win vs the result. Good at rating models like this
0 = perfect prediction (100% home win prob & home win)
1 = total wrong (100% home prob; away wins for example)
RPS works off of pre-match probabilities of a home win, draw, & away win vs the result. Good at rating models like this
0 = perfect prediction (100% home win prob & home win)
1 = total wrong (100% home prob; away wins for example)
September 6, 2025 at 1:24 PM
Here are the top 10 biggest upsets in T5 last season
RPS works off of pre-match probabilities of a home win, draw, & away win vs the result. Good at rating models like this
0 = perfect prediction (100% home win prob & home win)
1 = total wrong (100% home prob; away wins for example)
RPS works off of pre-match probabilities of a home win, draw, & away win vs the result. Good at rating models like this
0 = perfect prediction (100% home win prob & home win)
1 = total wrong (100% home prob; away wins for example)
Caveat: this was technically the #2 biggest upset of the season but the #1 was Bochum beating Bayern in Munich, but Bayern got an early red card and these models don't account for in-game state changes like that
September 6, 2025 at 1:24 PM
Caveat: this was technically the #2 biggest upset of the season but the #1 was Bochum beating Bayern in Munich, but Bayern got an early red card and these models don't account for in-game state changes like that
Finally spent the time to use my team rating model to create outcome probabilities instead of just an xGD
Compared it with the paper I worked off of & my model is at least as strong
In Top 5 leagues last season, the biggest "surprise" result was Sociedad 1-0 Barcelona
🧵
Compared it with the paper I worked off of & my model is at least as strong
In Top 5 leagues last season, the biggest "surprise" result was Sociedad 1-0 Barcelona
🧵
September 6, 2025 at 1:24 PM
Finally spent the time to use my team rating model to create outcome probabilities instead of just an xGD
Compared it with the paper I worked off of & my model is at least as strong
In Top 5 leagues last season, the biggest "surprise" result was Sociedad 1-0 Barcelona
🧵
Compared it with the paper I worked off of & my model is at least as strong
In Top 5 leagues last season, the biggest "surprise" result was Sociedad 1-0 Barcelona
🧵
Raw data graph for reference
September 5, 2025 at 4:43 PM
Raw data graph for reference
Trump can fire as many BLS directors as he wants but it won't change the fact that the 12-month slope for total non-farm employment has only been lower than it is now 2 times in the last 20 years
COVID, obviously
& between Sept 2007 - April 2011
COVID, obviously
& between Sept 2007 - April 2011
September 5, 2025 at 4:43 PM
Trump can fire as many BLS directors as he wants but it won't change the fact that the 12-month slope for total non-farm employment has only been lower than it is now 2 times in the last 20 years
COVID, obviously
& between Sept 2007 - April 2011
COVID, obviously
& between Sept 2007 - April 2011