RocketSolver — Easy-to-use next-generation poker solver.
Website: rocketsolver.ai
Pricing: $299 lifetime license for two devices (with one year of updates included).
Contacts: Discord — https://discord.gg/dwDu7Vr, Email — [email]support@rocketsolver.ai[/email]
Hello everyone!
I'm Yury, the developer of RocketSolver — next-generation desktop solver with a focus on usability and performance.
As a professional poker player in the past I always struggled with how unnecessarily complicated and annoying poker software can be. This product is an attempt to fix that once and for all. We designed it to be accessible to anyone, so you can focus on what really matters — winning at poker. RocketSolver also delivers exceptional performance while allowing you to solve complex situations without absurd hardware requirements.
MAIN FEATURES:
- Optimized for solving trees with up to 6 players starting from preflop
- Intuitive and easy-to-use tree building interface
- Preflop strategy locking is available, with postflop strategy locking coming soon
- Low memory requirements via using state-of-the-art abstractions to allow for solving extremely large trees
- Rake support
The product is in its early stages with many more features to come in the near future. We are committed to bringing the best possible poker solving experience to our users.
More info at rocketsolver.ai
If you have any questions, suggestions or feature requests, don't hesitate to leave a message here or contact us in our Discord or via [email]support@rocketsolver.ai[/email]
Tree building interface:
Strategy explorer:
Range editor:
11 Replies
how does the performance compare to monker, and how does the solutions compare?
does it scale in performance with higher end cpus?
some benchmarks would be good.
app does look nicer than monker.
how does the performance compare to monker, and how does the solutions compare?
does it scale in performance with higher end cpus?
some benchmarks would be good.
Hi,
It is extremely difficult to directly compare solutions between different solvers. For large multiway solutions, finding Nash distance is computationally intractable, so the only reasonable benchmark would be head-to-head performance. To do that, one would need to build a set of completely identical trees in both solvers (which can be challenging) and reverse-engineer the save file format to query the strategy. This is possible, but it would take a lot of time, so I haven't done that myself.
What I can say with reasonable certainty is that the core algorithm and abstraction techniques used in RocketSolver are more up-to-date and have been shown to generally produce stronger strategies than the Monker's approach, which is considered to be outdated. RocketSolver also has much lower hardware requirements, allowing it to solve larger trees.
The solving speed scales really well with CPU core count, and some additional work is being done on better utilizing instruction sets available on higher-end CPUs, so expect additional speed-ups in the future.
You are welcome to give RocketSolver a try by signing up for a free trial period. If you have any questions, please let me know. I'm more readily available on Discord and over email, but I will make sure to check this thread more often as well. Thank you for your interest!
Hey everyone! RocketSolver 1.9.0 is out now!
NEW FEATURES:
⦁ We've added support for incremental bet sizings. This feature allows for conveniently taking limpers or cold callers into account when adding preflop raises. For example, entering 2.5bb+ as RFI will automatically increase the final 2-bet size by 1bb for each limper. Using 3x+ as a 3-bet can spare you from having to specify the squeeze size manually. Long form is supported as well if you want to use a different increment, e.g., 9bb + 2.5bb.
⦁ Geometric bet sizing is now supported. You can now enter e to let the solver select the appropriate bet size automatically or specify the number of bets to split the stack into yourself, e.g., 2e.
⦁ You can now use the Min bet sizing to add a minbet or a minraise.
IMPROVEMENTS:
⦁ A new option is available to take rake preflop only if there was a 3-bet.
⦁ We've added turn and river abstractions of size 100 for preflop solutions to better accommodate solving complex full-ring games.
⦁ 2-card and 1-card BDFD categories have been added to the hand filter.
⦁ The "Convert bets to all-in" threshold setting has been split into two: one for preflop and one for postflop.
⦁ Bets are now uniformly displayed as big blinds on preflop and as a percentage of the pot on postflop, regardless of which bet sizing options were used to add them.
⦁ You can now use "bb" bet sizings on postflop.
⦁ Bet sizes less than the minimum allowed bet size are now automatically converted to the minimum allowed bet size when applied.
⦁ For Pot-Limit trees, bet sizes are automatically capped at pot bets when added.
⦁ Range import now supports the "plus" syntax ("66+", "A5o+").
⦁ A "Clear" button has been added to the "Select board cards" dialog.
⦁ More fields now support Tab navigation.
⦁ We've made additional UI rendering performance optimizations.
⦁ We've significantly reduced the number of cases where performance might throttle when the app is in the background.
FIXES:
⦁ Fixed an issue where the Numpad's Enter key could not be used to enter bet sizings.
⦁ Resolved an issue where the strategy matrix failed to properly account for dead cards when calculating cell height on postflop.
Best of luck at the tables!
Hi everyone! I'm happy to announce that RocketSolver version 1.10 is now available!
NEW FEATURES:
⦁ Panorama is a new feature allowing to view strategies across multiple preflop nodes simultaneously in a grid format. For example, when opened from the root node, it displays RFI strategies for all players. You can navigate between the nodes by clicking the actions in the average strategy bar.
⦁ The new "Bunching Ranges" feature now allows you to specify folding ranges for players who are no longer in the game, taking the so-called "card bunching effect" into account. This greatly simplifies the process compared to the previous need for tree pruning and node-locking.
⦁ New settings for "Max number of calls" and "Max number of limps" limit consecutive calls to drastically reduce the size of large multiway trees. Calls that close the action are always allowed.
IMPROVEMENTS:
⦁ The main window is now resizable.
⦁ Increased the maximum stack size to 1000 bb.
⦁ Added a turn abstraction with a size of 2000 for flop solutions.
⦁ The IP/OOP bet sizing filter is now available for postflop.
⦁ The tree population process can now be canceled in progress.
⦁ Improved handling of out-of-memory errors when saving solutions.
⦁ The "Lock strategy" dialog's strategy bar now shows corresponding actions on hover.
⦁ A button that opens saved ranges in Windows File Explorer has been added to the range input dialogs.
⦁ An "Invert range" button (with a Ctrl+I shortcut) has been added to the range input dialogs.
⦁ Ctrl+E has been added as a shortcut for clearing the range in the range input dialogs.
⦁ It is now impossible to accidentally create a postflop solution with a 0 pot.
⦁ In the "Create new solution" dialog, the dead money and pot inputs no longer share values.
⦁ Improved and added tooltips, alongside smaller visual improvements and performance optimizations.
FIXES:
⦁ With the main window now being resizable, a longstanding issue where part of the window would not fit on the screen due to Windows display scaling settings should now be fixed. If you're still encountering problems, please let us know.
⦁ Fixed a bug that sometimes caused the preflop relative position filter to be applied incorrectly.
⦁ Fixed a bug that could result in squeeze bet sizings leaking into other spots.
We've also recently published a short Getting Started tutorial available at https://rocketsolver.ai/docs/introductio...
With this update, we've completed a major internal rework, paving the way for many new features and engine optimizations. Upcoming priorities include an Entropy rework, an optimized 2-player algorithm, scripting support, and aggregate reports.
Hi - I came across your solver looking for an alternative solver with a better price. Have been playing around with it and it seems to produce similar results for preflop solutions compared to Simple Preflop ... which I think is a good thing!
However, I notice some of the postflop outputs don't make sense ... like below it says to check/jam sometimes with a set in SBvBTN on this board ... and then for the BTN to call only with JTs (should never call that if SB always has a set). The solution from GTO+ is quite different as to which hands would be a check/jam as well.
Any idea when Trainer function coming?
Any sales, discounts?
Hi - I came across your solver looking for an alternative solver with a better price. Have been playing around with it and it seems to produce similar results for preflop solutions compared to Simple Preflop ... which I think is a good thing!
However, I notice some of the postflop outputs don't make sense ... like below it says to check/jam sometimes with a set in SBvBTN on this board ... and then for the BTN to call only with JTs (should never call that if SB always has a set). The solution from
Hi,
I'm very sorry for the extremely late reply. For some reason, I don't get email notifications for new posts in this thread. Please feel free to contact me at [email]support@rocketsolver.ai[/email] or on our Discord community server, and I'll be more than happy to extend your trial period.
Regarding your question: Do I understand correctly that you're browsing postflop strategy in a preflop solution? If yes, the postflop strategy won't be as precise as it would be in dedicated postflop solvers like GTO+ or Pio. It's computationally infeasible to solve all possible postflop spots on all possible boards to a high degree of accuracy in a single preflop simulation. The GTO+ flop solution considers only a single specific spot while RocketSolver preflop solution has to solve millions.
RocketSolver uses a technique called Imperfect-Recall Abstractions to cluster similar postflop hands together, making it possible to solve large multiway preflop trees. This results in high strategy convergence speed on the initial street, but at the cost of reduced accuracy on later streets. The main takeaway here is that the strategy for the initial street will always be as accurate as possible, while the strategy for later streets will not be as precise. If you want to get a high-accuracy flop strategy, you can create a dedicated flop solution in RocketSolver. In that case the flop will be solved without abstractions. In the future, I hope to add a feature that will allow resolving nodes to a high degree of accuracy while browsing a solution.
Let me know if you have any other questions. I promise to monitor this thread more closely 😀
Hi,
I'm very sorry for the extremely late reply. For some reason, I don't get email notifications for new posts in this thread. Please feel free to contact me at [email]support@rocketsolver.ai[/email] or on our Discord community server, and I'll be more than happy to extend your trial period.
Regarding your question: Do I understand correctly that you're browsing postflop strategy in a preflop solution? If yes, the postflop strategy won't be as precise as it would be in dedicated postflop solvers
Thank you! No big deal with the late response. This makes sense to me.
Hey everyone! RocketSolver 1.11.0 is now available for download!
NEW FEATURES:
⦁ Postflop node-locking is finally here. You can now lock strategies on the initial street in your flop/turn/river solutions.
⦁ Added support for the flat drop rake structure. This allows you to take the jackpot rake into account or better simulate your offline games.
⦁ A "Clear tree" button has been added, allowing you to quickly reset the tree to its original state by removing all bets and raises.
IMPROVEMENTS:
⦁ The node-locking UI has undergone a major overhaul:
- The "Lock strategy" button now opens a panel on the right instead of a modal window.
- You can use Space to enter the node-locking UI and switch between its modes.
- You can use Ctrl+C and Ctrl+V without entering the node-locking UI to copy and paste full node strategies between RocketSolver instances.
- You can now brush hands in the matrix and strategy table without selecting them first. Hold Ctrl to lock the current strategy instead of the custom one. Right-click to clear the lock.
- You can now undo changes with Ctrl+Z.
- Holding Alt during node-locking will show the original strategy.
⦁ Entropy formula has been reworked. The range of values is now more stable across various trees and abstraction options. Please see the docs or the entropy tooltip for reference values.
⦁ A strategy bar displaying the total strategy has been added to the Strategy tab.
⦁ Added a "Call to all-in threshold" setting that can help avoid unnecessary cold calls in multiway trees with asymmetric stack sizes.
⦁ Additional optimization opportunities were found, resulting in a 5-10% average solver speed increase.
⦁ Applying bet sizes is now ~2x faster.
⦁ Rake settings are now cached.
⦁ Improved UI responsiveness for better support of large screen resolutions.
⦁ UI rendering has been reworked to use less system resources during solving.
FIXES:
⦁ Fixed a bug where incomplete all-ins could sometimes reopen the betting action.
⦁ Fixed an issue where bets were sometimes not properly converted to all-ins in spots with asymmetric stack sizes.
⦁ Fixed range folder names displaying incorrectly when containing dots in their name.
⦁ Fixed a rare bet size formatting issue.
This update includes multiple quality-of-life improvements and fixes that I would like all our users to have. I've made our early adopters with expired licenses eligible for this update without requiring a license renewal.
I'm continuing my work on integrating a specialized HU algorithm into RocketSolver, which I hope to release with the next update. I welcome your feedback and wish you good luck at the tables!
Hello! I am only able to nodelock strategy on the flop. Once I get to the turn, the `lock strat`` button becomes unclickable... Its about a 3way spot that remained 3 way till the turn. Any tips?