RangeSharp
RangeSharp
A first-person account of 30 days training preflop ranges. BTN accuracy from 62% to 94%. What worked and what didn't.
I was skeptical. Spaced repetition for poker ranges? Sounded like Anki for flashcards. I'd tried memorizing ranges before. Print a chart, stare at it, forget it by Tuesday. I figured this would be the same thing with a fancier UI.
It wasn't. Here's what happened over 30 days.
I started with BTN Open Raise. Easiest spot, right? Everyone says the button is the money seat. Open wide. It's easy. I'd read the charts. I knew it was roughly 40%. I figured I had a decent mental map. Last to act, open 40% of hands. How hard could it be?
Pretty hard. First session, 50 hands, 62% accuracy. I kept folding K9s. I kept raising 72o. The suited connector line was a mess. Where does 65s start? 76s? I was guessing. The system showed me the correct answer each time. I felt like an idiot.
I also noticed the UI. Decision mode deals you a hand, shows a poker table with your position highlighted, and you pick fold, call, or raise. Keyboard shortcuts: 1 for fold, 2 for call, 3 for raise. Space to advance. No mouse needed. I got into a rhythm. Wrong, see answer, next. Wrong, see answer, next. The FSRS algorithm was logging everything. I had no idea what it was doing with that data yet.
I stuck with it. Twenty minutes a day. BTN Open only. By day 5, I'd stopped folding K9s. By day 7, I was at 78% on BTN. The spaced repetition was working. Hands I'd missed kept coming back. I got them wrong again. They came back sooner. I got them right. They drifted to longer intervals. Some hands started feeling automatic. AA, KK, 72o. Obvious. The edges were still blurry: K8s, Q9s, 54s. Those showed up more often.
I added CO Open and HJ Open. Smaller ranges. Tighter. I ran 30 hands on each. CO: 58%. HJ: 61%. I was folding suited aces I should have raised. A8s, A7s. I kept thinking "that's too loose" and folding. The solver said raise. I was wrong.
Recall mode is different. No hand dealt. You paint the entire range on the grid from memory. I tried it on BTN Open. I thought I had it. I'd been drilling for two weeks.
I got 71% of the cells correct. I missed the bottom of the suited connector range. I over-included some offsuit junk (97o, 87o) and under-included suited gappers. The shape of the range was roughly right. The edges were wrong. Recall mode doesn't let you hide. You have to produce the whole thing. I'd been recognizing correct answers in Decision mode. Producing them from nothing was harder.
I ran Recall once a week after that. It kept me honest.
I added 3-bet spots: BTN Open vs BB 3Bet, CO Open vs BTN 3Bet. Fold, call, 4-bet. Mixed strategies in some spots. The system supports them. You see "75% 4-bet, 25% call" for a hand. I had to get used to that. The RNG picks an action when both are correct. Fair enough.
BTN Open was now at 88%. CO at 82%. HJ at 79%. The early spots felt different. When I got dealt 76s on the button, I didn't think. I raised. When I got dealt A5s in the cutoff, I raised. The algorithm had shown me those hands enough times. They'd moved to longer intervals. I wasn't reviewing them every day. I was reviewing them every few days. And I was getting them right.
The borderline hands were still the problem. K9s in position. QJo from the CO when UTG had opened. Suited aces from the BB when the button had 3-bet. Those kept appearing. I was improving, but slowly.
BTN Open: 94% accuracy. CO: 88%. HJ: 86%. I'd stopped second-guessing the obvious hands. AA from the button: raise. 72o: fold. The system had drilled those into me. The improvement was in the edges. K8s, Q9s, 54s. Hands I used to pause on. Now they were automatic. I'd added BB Defense vs BTN and BB Defense vs CO. Those were at 81% and 78%. Eight spots total. Mastery level 3 on most of them (the dashboard shows 0-4; 3 means "solid, some review needed").
The streak widget was satisfying. I had a 12-day streak at one point. Missing a day reset it. That's shallow motivation, but it worked. I didn't want to break the streak.
The borderline hands. K9s, suited aces, small pocket pairs. The hands that are sometimes in, sometimes out, depending on the spot. I'd memorize one context and mess up another. K9s from the BTN: raise. K9s from the BB vs a CO 3-bet: fold. The algorithm kept drilling those. I got better, but they were the last to stick.
Suited connectors were another leak. I'd over-fold them early (thinking they're too weak) and under-fold them in bad spots (thinking they're always playable). The AI debrief caught this. After a rough BB vs BTN session, I got: "You tend to fold suited connectors that have good equity vs the BTN's range. Consider drilling A5s-A9s and 65s-98s in this spot." I did. Accuracy on that spot went from 72% to 81% in a week.
There was a moment around day 18. I got dealt 54s on the button. A few weeks earlier, I'd have hesitated. Is that a raise? I'd have guessed. This time, I raised without thinking. The hand had come up maybe eight or ten times. I'd gotten it wrong twice. Then right. Then it had faded into the background. When it reappeared, I knew it. No recall effort. Automatic.
That's when I believed the method. Not because of the numbers. Because of the feeling. The hand was just... in there.
I ran a session on BB vs CO 3-bet. 68% accuracy. Bad. I hit "Get AI Debrief." The analysis said: "You're over-calling with broadway hands that are dominated (KJo, QTo) and under-calling with suited aces and suited connectors that realize well." I hadn't seen that pattern myself. I'd been reviewing mistakes one by one. The debrief connected them. I focused on those hand classes. Next session: 79%.
That's the value. Not "you got KJo wrong." "You're over-calling with hands that get dominated." Fix the concept, and the individual hands improve.
I started skeptical. I thought it was flashcards with extra steps. I committed because I had nothing to lose. Twenty minutes a day. Let's see what happens.
By day 30, I had a structure. Eight spots. Ranges I could recall. Accuracy in the high 80s and 90s on the spots I'd trained longest. The method works. It's not magic. It's repeated recall at the right intervals. Your brain does the rest.
I'd add Recall mode earlier. I waited until day 14. I should have tried it on day 7. It forces you to produce the range, not just recognize it. That distinction matters. Recognition is easier. Production is what you need at the table.
I'd also run the AI debrief more often. I only used it a few times. When I did, it pointed at leaks I hadn't noticed. More debriefs would have accelerated the fix.
And I'd pick one hand class to focus on per week. Week 1: suited connectors. Week 2: suited aces. Week 3: small pocket pairs. Week 4: marginal broadways. I spread myself across spots. Focusing on one hand class across multiple spots might have been more efficient.
The analytics dashboard became useful around day 10. Streak widget: trivial but motivating. Accuracy trends: I could see BTN improving week over week. Weak spots: the system highlighted BB vs CO 3-bet when I was at 72%. I drilled it. Today's focus: it suggested spots that were due for review. Recent activity: a log of what I'd trained. Nothing fancy. Just enough to feel like I was making progress.
The mastery levels (0-4 per spot) gave me a target. Level 3 felt achievable. Level 4 felt like "I really know this." I hit level 3 on BTN Open by day 25. Still working on level 4.
Thirty days. Eight spots. BTN Open 94%, CO 88%, HJ 86%, BB vs BTN 81%, BB vs CO 78%. The rest in the high 70s. Total training time: roughly 10 hours over 30 days. About 20 minutes per day. No marathon sessions. No cramming. Just consistency.
If you're on the fence, try it for two weeks. One spot. BTN Open. Twenty minutes a day. See if the hands start sticking. I think they will.
Master your preflop ranges.