I am not a disinterested observer of SmashIQ. I built it. I am also the primary test subject — every feature, every piece of feedback, every edge case I have shipped has been run against my own match recordings first. That means 200+ hours of my own sessions analysed, corrected, and re-analysed over two years.
This is my honest account of what changed in my game, what surprised me, and — importantly — where the AI hits a wall.
Where I Started
Two years ago I was a solid P3 player by any objective measure. Good enough to hold my own in club games, good enough not to be a liability to my partners. But I had the classic P3 ceiling: I was playing from the back of the court too much, my net game was reactive rather than active, and my bandeja technique was functional without being efficient.
I did not know any of this precisely. I suspected I was "not going to the net enough" because my coach had told me that. But I did not know the numbers.
Month 1–3: The Humiliation Phase
The first thing SmashIQ showed me was a volley rate of 11%. That is bad for a P3 player. The data from our broader match corpus shows P3 players typically volley at 14–16%. I was below my own peer group by a meaningful margin.
More humbling: when I did volley, my first-volley-to-net success rate (did the ball land in the intended zone?) was 61%. The SmashIQ model was charitably calling a lot of my volleys "successful" by primary classification, but cross-referencing against my partner's likely defensive options, it was flagging many of them as "positional errors" — volleys that were technically in but gave the opponents an easy recovery.
This was useful, specific, actionable information. My coach had told me "go to the net." SmashIQ told me "your volley rate is 11%, your first-volley positional error rate is 39%, and 71% of those errors happen when you are volleying from behind the service line." Same message, wildly different precision.
Month 4–8: Drilling Against Real Data
Once I had the precise problem, I could design precise practice. I drilled net approach aggressively — not generic net approach drills, but specifically drills designed to get me comfortable volleying from inside the service line. I logged every session through SmashIQ.
By month eight, my volley rate was 19% and my positional error rate had fallen to 24%. Those are A1-territory numbers. But I had not advanced to A1 yet in practice — because my bandeja was still P3.
The Bandeja Problem
Improving the net game exposed the next bottleneck: when the opponent lobbed me — which happens more at A1 level, precisely because A1 opponents know how to punish forward players — my bandeja was the weakest part of the chain. SmashIQ's technique scoring module measures contact point consistency, follow-through arc, and landing zone accuracy for overhead shots. My consistency score was 58 out of 100.
This is where I started to understand something important about AI coaching: it is extraordinary at diagnosing what is wrong and roughly calibrating why. It is not a replacement for a human coach with eyes. I knew from SmashIQ that my bandeja contact point was late — I was catching the ball behind my head — but working out the kinetic chain reason for that required 40 minutes with a coach watching my body mechanics. The AI got me to the exact question. The human coach answered it.
Month 9–18: The Compound Effect
The most powerful thing about having 18 months of your own match data is the time series. I can look at my bandeja consistency score trending from 58 to 74 over the period when I was focusing on it. I can see my volley rate plateau at 19% for two months after a wrist injury before recovering. I can see exactly when I started using the chiquita under pressure — the shot does not appear in my shot distribution at all before month 12.
The data gives you accountability. It is very hard to fool yourself about whether you are improving when the numbers are sitting in front of you. This is something human coaches know about — consistency is 80% of improvement — but the AI version enforces it without judgement.
What the AI Cannot Do
- •It cannot tell you why your wrist is late on the bandeja. It can tell you your contact is late. The kinetic chain is a human coach problem.
- •It does not know when you are injured, tired, or having an off day. It reads every match the same way. You need to flag context manually.
- •It cannot replace a drill partner. Footage analysis tells you what — only repetitions tell your muscles how.
- •It has no emotional model. A good human coach reads when you need encouragement versus when you can handle hard truth. SmashIQ does not get tired of your mistakes, which is mostly good, but occasionally it surfaces the same issue for the 15th session running and you need a human to reframe it.
Where I Am Now
I am a mid-A1 player. My SmashIQ overall rating sits at 71/100 (our internal composite metric). I will not pretend the AI built that game by itself. But I genuinely believe I would still be a P3 ceiling player without the precision it gave me about what to fix and when. The feedback loop shortened by two years of confused grinding.
The honest version: SmashIQ is the diagnostic layer. Deliberate practice is still the engine. It just makes the practice vastly more targeted.