Telegram Sniper Bots Reviewed: BananaGun, Photon, Trojan, and the Fee-Drag Nobody Tells You About

It's 2 AM. A friend in Zurich shows me his phone. Photon is open, a chart is spiking vertically, and he's grinning: "Sniper bot, bro. $BONK-clone. I'm up 4x in eleven minutes."

Nice. I ask him one question. "How much did you deposit into the bot over the last month?"

He does the mental arithmetic. It takes a few seconds. "Maybe 12 SOL."

"And how much is the balance now?"

Longer pause. Scrolls. "6.2."

"So across everything you've traded this month, you're down half."

"Yeah but..." he waves the screen at me, "THIS ONE is 4x."

I've seen this conversation three times in the last six weeks. Always the same shape. One screenshot of a winner, a blurry memory of everything else, and a balance that says the story is not what it looks like. Every time, the bot user swears the app is printing money.

The bot is printing money. Just not for him.

So this week I sat down and did what I always do when the marketing and the math don't match. I rebuilt the fee calculation from scratch. Three of the biggest Telegram sniper bots (BananaGun, Photon, Trojan) all advertise roughly 1% per trade. I took $1'000 of simulated memecoin capital, ran it through a realistic round-trip on each bot, and added up everything that actually comes out of your wallet.

Spoiler: the real drag is not 1%. It's not even 2%. Depending on the bot and the network, it lands somewhere between 5% and 8% per round-trip. Every round-trip.

This article is the full math.

This is the third piece in our Bot Reviews series. Same methodology we used on Bybit Spot Grid and Bitget Copy Trading. Strip the marketing, show the math, let the numbers decide the verdict. No affiliate links. No rev-share. If we like something, we'll tell you. If we don't, same.

Telegram sniper bots are a different animal from grid bots or copy trading, so the article structure shifts. Sections 3 and 4 are the centerpiece: how these bots actually take your money, and a worked round-trip example so you can feel the drag in dollars, not percentages.

What These Bots Promise

Walk into any of the three product pages and the pitch is essentially identical. Big serif headline. A Telegram interface screenshot. An animated candle going vertical.

BananaGun (bananagun.io): "Snipe tokens the second they launch. Honeypot protection built in. Works on Ethereum, Solana, BSC, Base."

Photon (photon-sol.tinyastro.io): "The fastest Solana trading terminal. Sub-second execution. One click from discovery to position."

Trojan (Telegram bot + trojan.app): "The Solana terminal for serious traders. Copy trading, limit orders, sniper settings, all through Telegram."

All three show the same carousel of promises. Speed, honeypot protection, stop-loss/take-profit, copy-trade a wallet, anti-rug. All three quote "1% per successful trade." All three are transparent about the 1%, and all three neglect to mention the four other things that also come out of your wallet on every round-trip.

How They Really Work

Here's the part most users never think about.

When you tell BananaGun to "buy $500 of $PEPE," you don't interact with Ethereum yourself. You type a command into a Telegram chat. The bot, running on BananaGun's servers, holds a custodial wallet for you. That wallet was generated the first time you opened the bot. BananaGun's infrastructure holds the private key.

You can export the key. Most users never do.

The bot then submits a transaction from your custodial wallet to Uniswap (or Raydium on Solana, or the relevant DEX on whatever chain). It uses high priority fees, and on Solana, a Jito tip, to push your transaction to the front of the block. That's the "sniping" mechanic: they pay more to jump the queue.

Three things follow from this architecture.

First, your keys aren't in your wallet. They're on a server run by a team whose identity is partially anonymous. BananaGun itself was hacked in September 2024. 11 users lost around $3 million through a Telegram message-oracle vulnerability. The team refunded from treasury, which was the honorable move, but the architectural risk didn't go away. Your funds live in a hot wallet controlled by the bot's infrastructure. If it gets hacked, or if the team disappears, your SOL and your tokens are the team's problem, not yours.

Second, the "instant snipe" isn't magic. It's a priority-fee auction. On Ethereum, BananaGun pays a high priority fee to a validator or a private mempool. On Solana, all three bots pay a Jito tip, a bribe to the block builder to include your transaction first. The median retail Jito tip is around 0.1 SOL. During the TrumpCoin launch, the typical sniper paid a 1.2 SOL median and top traders 3.7 SOL per transaction. When a "1.2 SOL median tip" gets quoted during a launch event, that's real money leaving real wallets.

Third, the honeypot check takes time. BananaGun's simulator tries a dummy sell before confirming your buy. Photon offers static filter-based checks (mint authority, LP burned, dev blacklist) but no equivalent pre-trade sell simulation. Trojan has basic anti-rug filters but nothing as rigorous as Maestro's patented mempool-frontrun protection. The honeypot check is genuinely useful. The question is whether it's worth the fee drag to use a bot for it. We'll get there.

We Did the Math on the Fees

Here's the centerpiece. Three bots, one $1'000 memecoin round-trip, all fees totaled.

The Fee Schedules, What's Actually Published

Three bots, roughly the same 1% per side as of April 2026. That's 2% round-trip on a successful trade. So far, exactly what the marketing says. (Fee schedules can change. The structural argument that follows holds regardless.)

Now let's add the three things the marketing doesn't total up.

The Four Hidden Costs

1. Priority fee / Jito tip (Solana).

Retail users set tips between 0.001 and 0.005 SOL on normal trades. During contested launches, that jumps to 0.1-1.2 SOL per transaction. Using conservative retail defaults:

• Buy tip: 0.003 SOL, about $0.45 at $150/SOL
• Sell tip: 0.003 SOL, about $0.45
• Round-trip tip cost: ~$0.90 on low-contention trades

That's small on a quiet trade. But during a hot launch (the kind you're using a sniper bot for) it balloons. For a $1'000 buy with a 0.1 SOL tip (retail-average for active contention), the tip alone is $15 per side, $30 round-trip. That's 3% of your capital, gone, before any price movement.

2. Slippage on thin memecoin liquidity.

This is where most of the damage hides. On a brand-new pump.fun token with $40k of liquidity, buying $1'000 worth of tokens pushes the price up around 2.5%. Selling $1'000 worth pushes it down another 2.5%. That's 5% round-trip slippage on low-liquidity names. On a token with $10k liquidity (and snipers routinely trade these) it's more like 10%+ round-trip.

Default slippage tolerance on most sniper bots is preset to 10-15%. That's not because the team wants to rob you. It's because the trade would fail without that tolerance. You just pay the slippage instead. Same wallet, different column.

3. Failed transactions. You don't pay the bot fee, but you pay the gas.

All three bots are honest about this. They don't charge their 1% on failed transactions. Good. But you still pay the Solana network fee (~$0.0005, negligible) and, depending on how the bot structures the bundle, the Jito tip may be consumed even if the main transaction fails. Bundled-tip patterns (Jito's recommended flow) typically refund the tip on failure; non-bundled patterns don't. The bot you're using rarely tells you which pattern it's on. Multiple retail guides flag failed-tx priority-fee costs as one of the top silent wallet drains during memecoin launches.

During the April 2024 memecoin peak, Solana's non-vote transaction failure rate spiked above 75%. On a contested launch a retail sniper can easily eat several failed attempts before landing one, especially when slippage or priority tip is set conservatively. At 0.003 SOL per failed attempt, five failures is 0.015 SOL — another $2.25 of pure loss, with nothing to show for it. If tips are consumed on failure (see above — depends on the bot's bundle structure), a contested 0.1 SOL run could turn five failures into roughly $75. Gone, for the privilege of losing an auction.

4. MEV / bundle retention.

Here's the subtlest one. The Jito tip goes to the block builder to include your transaction first. But Jito also runs bundle auctions where MEV searchers can front-run or sandwich retail trades. The sniper bot teams have done what they can. BananaGun advertises multi-layer MEV protection and private transaction routing. On Solana in particular, though, the line between "protecting your trade" and "paying someone not to front-run you" is thin. For smaller bots (Photon, Trojan) without a private-mempool lane, a sandwich attack on a large market-order memecoin buy can cost 0.5-3% of trade value, with nothing on your statement showing it happened.

Worked Example: $1'000 Memecoin Round-Trip on Photon (Solana)

All numbers in USD, assuming SOL at $150, token at your target fill price.

Low-contention means a token that's been tradeable for a few days and has decent liquidity. Active-contention means you're sniping an actual launch. Which is what sniper bots are sold for.

In dollars: on a quiet trade, the bot takes ~8% of your capital on a round-trip. On a hot launch, ~20%. The advertised 1% is there. It's just one stripe of the total.

Comparison to Raw Jupiter / Uniswap

The reference point every sniper bot user should run in their head. If you'd made the same trade on Jupiter (Solana's leading DEX aggregator), directly from your own self-custodial wallet, you'd pay:

• Jupiter's aggregator fee: ~0.1% per side (usually absorbed into better routing prices)
• Solana network fee: ~$0.0005
• Slippage: same 3-7% (it's the market, not the venue)
• No bot platform fee
• No custodial wallet risk

Round-trip: ~6% on a hot launch, ~3% on a quiet trade. The fee delta to a sniper bot is roughly $50-140 per $1'000 round-trip. Per trade.

The question the sniper bots never ask: does the bot's speed edge create enough value to cover that $50-140 per-trade extra cost, reliably, across the base-rate of your trades?

We're going to answer it.

What the Math Says About Edge

Here's the clean version of the question. If the sniper bot's fee drag over Jupiter is roughly 5-12 percentage points per round-trip, the bot needs to front-run the average retail trader by at least 5-12 percentage points per round-trip to break even on the fee math alone.

Does it? Honest answer: rarely.

By Delphi Digital's measurement, BananaGun wins ~88% of top sniping bundles on Ethereum — meaning it dominates relative to other sniper bots competing for the same opportunities. Which sounds fantastic until you think about what it means. It means BananaGun beats the other bots. So do the hundreds of other sniper bot users launching at the same moment, just less often. You are not front-running retail. You are competing against the other 5'000 bots trying to snipe the same launch. Your edge over them is whatever the bot-vs-bot ranking says it is. Your edge over the retail user who manually clicked "buy" on Jupiter thirty seconds later? Maybe 5-15% on the fill price. Against the project insiders who seeded liquidity and know exactly when it drops? Strongly negative.

Now the base-rate question. Dune Analytics data on pump.fun traders is brutal:

• 60% of all memecoin traders lose money
• 99.6% of Pump.fun traders never realized more than $10'000 in profits
• Only 0.048% (1 in ~2'000) ever realized $100'000+
• 95.6% of wallets broke even or lost money

That's the pool the sniper bots are fishing in. Most participants lose. If the fee drag is 8% per round-trip and 60% of participants lose money before fees, the bot doesn't fix the base rate. It amplifies it. Add an 8-12% fee drag to a losing distribution, and more of the traders who would have broken even now lose too.

When It Might Actually Work / When It's a Tax on Hope

I want to be fair. These bots aren't scams, and there's a narrow scenario where the honeypot-protection feature alone is worth the drag.

The rare case where a sniper bot adds real value.

You're a highly-informed degen with a network of alpha groups, you personally know the devs behind three or four pre-launch tokens, you get signals 10-30 seconds before the public feed, and your winning trades are 10-50x. In that world, the bot's honeypot-check might save you from one rug per month, which easily pays for the fee drag across all your trades. The speed edge on genuinely alpha information is real and can be monetized.

This is roughly 1-3% of sniper bot users. The ones with private Discord access and actual non-public info. Not the 97% typing commands into a Telegram chat after seeing a token trending on DEXScreener.

The common case, tax on hope.

You follow crypto Twitter. You see a memecoin getting shilled. You open your sniper bot and throw $500-1'000 in. You get a near-the-top fill (because by the time the trend made Twitter, the real snipers had already exited). The coin either pumps 1-3x before you panic-sell, or dumps 50-80% and you're stuck holding it. Across 10-30 trades per month, the Dune base rate applies: 60% probability you lose, and the 8% round-trip fee drag accelerates that loss.

In Dune terms, if 60% of memecoin traders lose before accounting for sniper-bot fee drag, the net-of-fees number for active sniper-bot users has to be higher. We cannot verify the exact segmented figure — no public data separates "traded via bot" from "traded direct" in the pump.fun distribution. But the direction is unambiguous. Fees make a losing distribution worse, and sniper bot fees are not small.

The honeypot-protection argument, weighed honestly.

BananaGun's pre-trade simulation is genuinely good. If you'd otherwise trade a token that turns out to be a honeypot (where you can buy but not sell) the simulator saves you 100% of that trade's capital. A single honeypot avoided on a $500 trade saves $500. Fifty round-trips at 8% drag costs $200-400 per $1'000 traded.

So if your honeypot-encounter rate is above roughly 1 in 25 trades, the protection is arithmetically worth the drag. If it's lower (and for experienced degens using DEXScreener filters, it usually is) the drag outpaces the protection.

Verdict

I'm going to differentiate between the three bots, because they are not equivalent. Same business model, different trade-offs.

BananaGun, use with extreme caveats. The most feature-mature of the three as of April 2026. Honeypot protection is genuinely the best in the category, MEV routing is a real structural feature, the manual-swap 0.5% fee is half the sniper-bot category standard. The September 2024 hack is a real negative, but the team covered the loss from treasury and introduced post-hack security measures (2-hour transfer delay, 2FA on transfers — credibility win, not deal-breaker). If you're going to use a Telegram sniper bot, this is in our view the least-bad one. Fee drag is still 6-15% round-trip in real conditions. Not a recommend. An "if you must."

Photon, only if you specifically need the speed edge. Fastest Solana UI in the category and a clean interface. The trade-off: only basic static honeypot filters (no pre-trade sell simulation like BananaGun's), no MEV lane, and the 1% per side has to compete with using Jupiter directly from a self-custodial wallet at near-zero platform cost. We wouldn't use it for general memecoin trading because the speed edge is consumed by the fee drag on any trade we'd plausibly run. The narrow scenario where it earns its fee: sub-30-second snipes on freshly-launched tokens where every block matters and you're already accepting the fee drag as the cost of the speed advantage.

Trojan, we wouldn't use it — specifically because of the referral structure. The 5-layer-deep referral program is the feature the user sees last and the team profits from first. Up to 35% of referrals' trading fees flow upward to the upline (Trojan documents this themselves). BehindMLM has criticized this multi-tier structure and characterized it in pointed terms (linked review for the full quote). Whatever you call the structure, the incentive is clear: Trojan is rewarded more for recruiting new traders than for retaining profitable ones. That alignment is the opposite of what we want from a tool that takes a percentage of our trades. Every influencer-shilled sniper-bot link is, mechanically, a referral tree. We'd rather pay zero recruitment-incentive overhead and trade direct.

Will We Test This With Real Money?

No. The simulation answer is structurally clear, and the Dune base-rate data closes the door. Depositing $500 into a sniper bot to run 30 trades would cost us $150-400 in expected fee drag alone, and the experiment would confirm what the math already shows.

We might revisit if a sniper bot publishes the distribution of user outcomes (not aggregate platform volume), the ratio of winning to losing user wallets, and the median user's 90-day net return after all fees. None of the three bots publishes any of this. None of the competitor bots (Maestro, BONKbot, GMGN) publish it either. The silence is the answer.

Honest Disclosure

Affiliate relationship: None. BearBullRadar has zero affiliate links to BananaGun, Photon, or Trojan. Per our Bot Reviews editorial policy, we never take affiliate revenue from Telegram sniper bot products. The conflict is structural. The affiliate take is paid per-trade from the fees we just documented as the product's single biggest user-harm vector. We won't recommend something we'd rather people not buy.

Data sources cited in this review:

• Banana Gun fee schedule and hack disclosure (bananagun.io, Cointelegraph on Sept 2024 hack)
• Photon fee schedule (Photon Gitbook)
• Trojan fee schedule and referral structure (CoinCodex, Trojan docs, BehindMLM review, Wayback snapshot)
• Pump.fun / memecoin trader base rates (Dune Analytics via CoinMarketCap, Altcoin Buzz)
• Jito tip economics (alchemist1411 on Jito tips as MEV powerhouse, shamikhzafar0 on the anatomy of Jito tips during the TrumpCoin launch, Solana docs on fees)
• Jupiter DEX comparison (Nansen on Jupiter)
• Feature comparison (Maestro vs Trojan analysis, CoinGecko Telegram bot roundup)

Research limits worth knowing:

• Fee schedules can change. These numbers are accurate as of April 2026 based on the linked primary sources and multiple cross-references. A bot could lower its fee tomorrow. The structural argument about slippage + tip + failed-tx still holds.
• We did not make live trades. The round-trip cost table is built from the published schedules plus typical Solana network conditions. An actual live $1'000 round-trip on a hot launch could come in worse than our numbers, rarely better.
• The Dune statistics are for pump.fun specifically. Trojan, Photon, and BananaGun all route into Solana DEXs that overlap heavily with pump.fun liquidity but aren't identical. The direction (most traders lose) is robust across every segmented memecoin dataset we found.
• We cannot separate "traded via sniper bot" from "traded direct" in the aggregate Dune data. The segmented data doesn't exist publicly. If it did and contradicted our math, we'd update. We don't expect it to.

Related reading:

• Bybit Spot Grid: 1'961 Tests, first in this series
• Bitget Copy Trading: The Wall of Winners, second in this series
• Why Any Trading Strategy Must Beat HODL, the benchmark every review uses
• Bot Trading Explained, what bots do, what they don't

Coming next in the Bot Reviews series:

• Bybit Futures Grid Neutral: Monte Carlo on the liquidation probability nobody publishes.
• Pionex Infinity Grid: the "set and forget" promise tested across 4 market regimes.
• 3Commas Three Years After the API Leak: did they fix it?

Subscribe to the Bot-Letter to get each review the day it lands.