The Next Next: Update #31
Gearing up to hunker down.
For those of you tuning in for the first time, The Next Next is a ‘build in public’ type of expedition from founder Jason Jacobs (me!) to chronicle my path building a new kind of athlete development company from the ground up, starting with ice hockey.
I am learning in public, ideating in public, and eventually building in public as we try to create a new kind of development experience for athletes. One that’s digital-first, built for modern families, grounded in real performance improvement, and accessible to anyone, anywhere.
I am still super early, and as anyone who has built companies from zero before knows, what comes out the other side will likely look far different from where I am starting. But the benefits of building in public along the way are invaluable for accountability, feedback, accelerated learning, and meeting lots of interesting and relevant people and potential collaborators along the way.
Each week in the newsletter (subscribe here) I share behind-the-scenes updates: what we’re testing, what we’re hearing from the market, and what’s coming next. The podcast goes deeper, featuring conversations with coaches, founders, athletes, and experts who are shaping the future of sports and human performance.
If you want to catch up, the historical weekly updates are here:
Week 1, Week 2, Week 3, Week 4, Week 5, Week 6, Week 7, Week 8, Week 9, Week 10, Week 11, Week 12, Week 13, Week 14, Week 15, Week 16, Week 17, Week 18, Week 19, Week 20, Week 21, Week 22, Week 23, Week 24, Week 25, Week 26, Week 27, Week 28, Week 29, Week 30
Thanks for coming along for the ride!
______
Hi all,
Another productive week!
Refresher
As a reminder, I am building a new kind of athlete development platform, starting in hockey. The thought is that there is tremendous benefit to high end player development coaches who do film review and then personalize development plans for emerging players. But they are super expensive, and as a result, they really only make sense if your player is ultra serious or if you are ultra wealthy.
With the rate that computer vision/AI/machine learning are improving, I am setting out to see if we can teach the machines to do this kind of high end player development, starting with film review and individual player analysis. I don’t believe the machines will ever fully replace the humans, but I do think tech will enable the humans to serve a lot more people over time, meaning the price of these services per player/family could come down, increasing the accessibility of high end player development coaching to more players.
I have been coming across some great platforms focused on coaches/teams, but haven’t seen any that put the player/family at the center and provide the analysis/development plans too, not just the data.*
*I am learning more every week, so just because I haven’t come across it doesn’t mean it doesn’t exist
Progress this week
What I have been hearing from high end player development coaches is the following:
As their mastery grows, their client base shrinks as they serve increasingly high-end players, and they want to find ways to get their philosophies/learnings out to more players without diluting their approaches.
Many are one person operations, meaning they only get paid when they are the ones doing the work. They see value in having at least a leg of their business that keeps running whether they are doing the work or not.
In order to climb the ranks in that profession it often involves traveling a ton, and some are seeking ways to take on new challenges while still being present with their young families.
As a result, I have found several that would be keen to see if we can get something going here, and I am just beginning to make the rounds.
Status on the tech side
This was the first week that I started reaching out to tech people in earnest, to try to figure out a gameplan for how this could get built, what order to do things, where to start, what kind of skills I need around the table, etc.
First of all, teaching machines to watch a hockey game and have insights about how an individual player performs is super hard. It isn’t like analyzing a golf swing or counting free throws. It involves a bunch of non-obvious insights that might not come out in the data (how does a player move without the puck, how quick/sound is their decision-making, how quick are their first 3 strides when they change direction, etc). Here is a clip from a data scientist for the Chicago Blackhawks on why hockey is so hard for AI to figure out.
That being said, tech is improving rapidly under our feet (just look at what 49ing is doing!), and many experts are telling me if you are smart about how you go about it, you could make a lot more headway in this area than most people currently believe. Personally, I am attracted to seemingly insurmountable problems like this one that would be a huge deal if solved in a compelling way. The kind that, even if you fail, if you helped move the category forward in the process would be worth it.
I have a number of calls scheduled w/ PhDs, professors, AI/machine learning engineers, founders of computer vision companies, etc etc to start to judo this area and flesh out our gameplan. Ideally, as I home in on the skills we need through these discussions, I will home in on the right person to work with in this capacity as well. It is a hugely important role for us, as I am now confident we have the right player development people around the table so once we have a tech gameplan laid out I am feeling ready to rock.
Competitive landscape
There seem to be a number of freelance video review coaches offering their services. It feels like the companies like Sportlogiq and 49ing that currently are building AI-powered solutions for teams/coaches may aspire to move towards individual player development as well. And of course there are the sensor-based approaches from companies like Helios and Catapult. I have deep respect for all of these companies, and don’t aim to reinvent the wheel in any capacity, if I can avoid it. For those that want to explore partnering, I would welcome those discussions. If there are foundational elements we find we’ll need to power our offering, I will only recreate them if left with no other choice. And we could be a channel to help sell your services as well, especially as we start to get some scale (if we do, of course!).
Outstanding questions
I have a lot of outstanding questions!
Here are a few:
We can get the shifts clipped easily from Sportlogiq/Livebarn one game at a time, but that doesn’t scale. Will either need access to an API from them, or will need to find another way to get the game footage.
Sportlogiq/Instat/49ing are another good source of this game footage, but they only seem to offer expensive licenses to teams. Would be interested in partnering, as we don’t seek to recreate what they have done. Our ideal world would only focus on the analysis and personal development plans, and leave the data collection/annotating to the experts.
The processes of these player development coaches today are highly manual, so it will be interesting to start to pick off key aspects of their work where the fruit might be ripe for the machines to help that are compelling enough to provide standalone value but narrow enough to get to market quickly. Determining which these are seems like a non-trivial task.
There doesn’t seem to be any platform purpose-built for player development coaches to work with their private clients. This is not the direction we are planning to pursue, but maybe an opportunity for someone.
My focus for the next week
The priority is absolutely starting to push hard on the technical front, in terms of scoping out our approach, staging, resources, etc. It is time!
Some reading material
A few articles I was sent this week that I am eager to get through, and maybe you will be too:
Here’s an article on AlphaGo Zero (the model trained using reinforcement learning): https://deepmind.google/discover/blog/alphago-zero-starting-from-scratch/
Here’s an introductory article from Hugging Face re: RLHF: https://huggingface.co/blog/rlhf
Here’s a good intro to RL from Towards Data Science: https://towardsdatascience.com/introduction-to-reinforcement-learning-c99c8c0720ef/
Here’s a talk on sports analytics and why it’s a great fit for AI:
Here’s a recently published player reidentification model by sportsSUSHI that was specifically trained on a new hockey dataset:
https://arxiv.org/abs/2502.21242v1
How you can help
Advice on how to train AI models around player tracking/coaching
Intros to people w/ these skills that may have a passion for startups and/or hockey
Connections to more player development coaches, team coaches, agents and scouts, as well as high-end practitioners in adjacent areas serving the same demographic like strength, mental performance, and nutrition.
Connections to founders building companies in relevant areas, like computer vision applied to other categories, player tracking in other sports, etc.
Correct me on any false assumptions I made anywhere above, as I am still getting up to speed, and doing my best!
New content this week
Two new episodes this week!
One with the two SportsRecruits co-founders (recently acquired by IMG Academy), Chris Meade and Matt Wheeler. apple. spotify. youtube.
And one with IMG Academy mental performance coach, John Couture. apple. spotify. youtube.
I think that’s all, have a great week! Thanks to all the people who have been helping with intros/ideas/encouragement, it is tremendously appreciated.
Jason