Session log: 135 rounds of Tower Rush 2

Both sessions in one view

Two sessions, two very different returns — and both inside what a 97% RTP game can do over a few dozen rounds. The second session is the better guide to the Auto-Bet strategy because it is four times longer; the first is the better guide to the manual exits because it is the one where we used them.

Session 1 — 35 rounds, phone client

Rounds 1–10 — manual, floor by floor

Floors are listed in landing order. "Peak total" is the product of landed floors before the collapse or cashout.

Rounds 11–35 — Auto-Bet, 3.00× stop, 2X Chance on

Each round cost 125 (100 + 25% ante). Auto-Bet cashed out the moment the total reached 3.00×, so winning rounds paid between 3.07× and 4.60× of the base bet.

All 40 landed floors of session 1

0.9 · 1 · 2.5 · 0.45 · 0.9 · 2.35 · 2.85 · 1.05 · 1.1 · 1.4 · 2.8 · 1.05 · 1.1 · 1.6 · 1.1 · 2.2 · 0.4 · 1.1 · 2 · 0.75 · 0.45 · 2.55 · 1.45 · 2.55 · 1.3 · 1 · 0.75 · 1.65 · 0.8 · 1.5 · 0.5 · 2.7 · 0.8 · 0.6 · 2 · 0.5 · 1.5 · 2.4 · 1.35 · 0.4

Progress Bar in session 1

Empty at the start; about one third full after the 36th bet. Twenty-five of those bets carried the 2X Chance surcharge, which does not change bar speed. No Bonus floor landed and no Triple Build floor appeared in these 35 rounds.

Buy Bonus purchase (after round 35)

With the bar a third full we bought the Bonus Round for 10,000 (100× the 100 stake). Ten drops later the result card read 83.35× — 8,334.95 paid on the base bet, a net −1,665.05 on the purchase. It is counted separately from the session-1 economics above.

Session 2 — 100 Auto-Bet rounds, desktop client

Same day, fresh 100,000 balance, the landscape desktop client. Stake 100, Auto-Bet set to 100 rounds with "cash out by total coefficient" at 3.00, 2X Chance off, no manual input from the first BUILD to the last. A script copied the Results column and the balance as text every 0.4 s; outcomes come from the History tab. The window was hidden twice for a few minutes early on, which froze the client (it does not run in a background tab) — the floor sequences of 24 rounds in that stretch are missing, their outcomes are not.

The 17 winning rounds

"Total" is the multiplier at the moment Auto-Bet cashed out. Anything well above 3.00× is a round where the floor that crossed the threshold was large, or where a Triple Build dropped three floors at once and the stop rule fired only after the third.

76 rounds, floor by floor

Landing order, left to right. Product is the multiplier of the whole tower when the round ended — by cash-out (bold) or by the crane missing. Note how often a tower of ten or more floors ends below 1×: a long round is not a growing round.

Distribution of the 346 floors

The seven floors at 3× or more were 3.0, 3.0, 4.0, 4.5, 4.5, 5.0 and 5.0. Session 1 never produced one in 40 floors; session 2 produced one every 49 floors. The share of sub-1× floors also moved, from a third to a half. Both are what you expect from samples this size, and the reason we publish the raw sequences rather than a single average.

Free Bonus Round — 1.00×

On the 100th round a Bonus floor landed on the tower while the total stood at 1.31× (131.33 FUN). Auto-Bet stopped, the night tower opened and ten floors fell: 0 · 0 · 1.9 · 0 · 0 · 1.7 · 0 · 0 · 0 · 0. Eight of them were x0. The 1.9 was erased by the x0 that followed it, then the 1.7 by its follower, and the card read 1.00× — the 131.33 held at the trigger was paid and nothing more. It is the worst result a Bonus Round can give and a useful one to have on record: "cannot bust" means the stake is safe, not that the round will pay.

Progress Bar in session 2

Empty at the start, roughly three quarters full after round 88 (no 2X Chance this time). It had not reached the top when the Bonus floor arrived, so this session says nothing about how many 100-stake bets fill it; the first session's "about a third after 36" remains the only data point.

Method and caveats

• Floor values were read from the Results column as page text immediately after each drop (by hand in session 1, by a 0.4-second polling script in session 2); outcomes from the History tab; balances from the header.
• Session 2: the sum of the 17 History wins (9,851.69) equals the balance change plus the stake exactly, which is our check that no round was missed. Floor sequences are missing for 24 rounds during which the client was frozen in a hidden window; those rounds all appear in the outcome table.
• Rounds 5–7 collapsed on the first floor, so they contribute no floor data.
• One 50% Cashout (round 3) and one full cashout (round 10) were taken deliberately to document both exits; all other manual rounds were built until collapse, which lowers the return versus a cashout strategy.
• 135 rounds is still a small sample. It cannot confirm or contradict the stated 97% RTP; it can only show the texture of a session.

Vikram Sethi

Lead reviewer · Bengaluru

Vikram spent seven years as a QA engineer at two mobile game studios before switching to independent reviewing in 2021. He tests every fast game the same way: a fixed stake, a written protocol, and a round-by-round log pulled from the game's own history panel. He does not accept review fees from operators.

How we testFull session logReport an error

Fact-checked by: Sana Qureshi — Fact-check · LahorePrimary sources: in-game "How to Play" text (Galaxsys client), galaxsys.co game page, our own demo history exportLast verified: 2026-11-10