Grid vs DCA bots: what changes in a range, rally or sell-off?

What is the difference between grid and DCA crypto bots?

A grid makes repeated buys and sells at specified price levels. A price-triggered DCA bot adds a predefined order after a decline and may exit the combined position at a target. Calendar DCA invests a fixed amount on scheduled dates whether price rises or falls; it is a different policy.

The important distinction is inventory behavior. Grid profits often recognize completed low-buy/high-sell pairs while the remaining inventory can fall. Averaging lowers average purchase price by spending additional cash; it does not recover a loss merely because that average becomes lower.

Before comparing results, obtain each bot’s range, order sizes, reinvestment and exit rules. A platform using the same “DCA” name can implement different conditions. Our illustrations below use one stated long-only spot policy for each category and no leverage.

Verify: 3Commas grid setting mechanics · 3Commas price-triggered DCA settings · Fixed-date BTC DCA is a different study

What exact rules and capital do the three examples use?

Every hypothetical account starts with $1,000 and a reference token at $100. The grid buys five units initially, keeps $500 cash, buys one unit at each first downward crossing of $90 and $80, and resells those added units on the next upward grid step. Original inventory sells one unit at $110 and one at $120. No trailing, expansion or re-entry above $120 is allowed.

The averaging bot buys $200 at $100, another $200 only on the first visit to $90 and another $200 only on the first visit to $80. It does not sell in these short examples. A buy-and-hold reference buys ten units at $100; cash stays uninvested. We assume exact synthetic level fills, no spread, tax or financing; the before-cost result and a separate 0.10%-per-fill cash-fee sensitivity are both shown.

Three paths are frozen: range 100→90→100→90→100; rally 100→110→120→140; sell-off 100→90→80→70. These invented paths isolate mechanisms. They are neither empirical BTC observations nor probabilities for tomorrow’s market.

Can the same grid make money in a range but lose in a sell-off?

Yes, in this hypothetical ledger. The range completes two $90→$100 one-unit cycles, realizing $20, while five units remain worth $500. In the sell-off, there are no completed sales: cash is $330 and seven units are worth $490, leaving $820. Inventory loss is an account loss even when a platform emphasizes its completed cycles.

The rally shows the opposite trade-off: selling inventory can reduce participation in further upside. At $140 the grid holds three units and $730 cash, so equity is $1,150; the fully invested reference is $1,400. That difference comes from exposure and specified exits, not a universal claim that one bot is superior.

The table keeps all three paths and both policies, including losses and unspent cash. Values are gross marked equity, not liquidated proceeds. Selling remaining inventory would add another fee and an actual bid/ask execution question.

How do you reconcile bot profit with total account equity?

For a fully paid spot account, equity equals remaining cash plus token quantity times the marking price. It is not cash plus inventory plus realized profit: realized proceeds already changed cash. Adding the profit counter again double-counts it.

Use the range-grid example: $500 initial cash − $180 buys + $200 sales = $520 cash. Five original units at $100 add $500, yielding $1,020, not $1,040. For the falling grid, total inventory acquisition cost is $670 and marked inventory is $490; the $180 difference explains the account decline.

For a real bot, reconcile starting and ending balances with deposits, withdrawals, fees and every fill. Use the same reporting currency for all inventories. A futures account needs a separate cash/realized/unrealized/margin calculation; this spot identity must not be reused as a futures liquidation model.

Verify: Read the actual ledger of a claimed bot result

When is a grid step too small to cover fees and execution costs?

A completed pair must cover both fills and execution loss, not only one fee. For one unit bought at $100 and sold at $101 with a 0.10% fee on each notional, gross gain is $1 and fees are $0.201; net gain before spread or slippage is $0.799. Buying at $100 and selling at $100.10 gives only $0.10 gross against $0.2001 fees, a $0.1001 loss.

For the frozen three examples, cash-paid fees reduce grid equity to $1,019.12, $1,149.27 and $819.33 in range/rally/sell-off. Averaging equity becomes $1,021.82, $1,079.80 and $869.96. These are marked accounts with fees on actual listed fills only, not a final liquidation fee. A fully funded hold reference buys 9.99000999 units at 100, paying its entry fee within the same $1,000 budget; marked equity ends at $999.00, $1,398.60 or $699.30. No account borrows cash to pay this illustration’s fees.

Use actual order prices when they already embed spread or slippage; do not add the same cost again. A limit order is not necessarily fee-free or filled. Queue position, thin books and partial fills can change both realized cycles and inventory.

Verify: Trading-fee break-even calculator and units

How much capital do averaging orders and multiple bot pairs really require?

Budget every potential order, not just the first deal. A hypothetical $100 initial order plus four averaging orders of $100, $200, $400 and $800 requires $1,600 before fees; two simultaneously active pairs require $3,200. Reserving only the first $100 can cause the bot to stop exactly when its advertised averaging rule needs more funds.

Publish the maximum inventory, outstanding reserved orders, cash buffer and total correlated-pair exposure. Several altcoin bots can consume that reserve together during a BTC sell-off. “Only one bot” does not mean one position when it monitors many pairs.

Trailing moves a range under stated rules; expansion adds levels and may require additional capital. Neither removes existing inventory risk. The vendor’s documented stop action must specify whether it merely stops new orders or also sells assets.

Verify: Grid trailing, expansion and stopping actions · DCA total order budget and multiple pairs

How would you fairly backtest grid versus DCA in historical markets?

Freeze a shared opportunity calendar, initial cash, asset eligibility, orders, costs, marking and final treatment of inventory. Compare cash, buy-and-hold, grid and the explicit averaging policy on the same periods. Classify a market state using only earlier completed data, not the later path you want a strategy to exploit.

Record net equity, drawdown, inventory concentration, reserve use, time underwater, unfilled orders and terminal liquidation sensitivity. Include transitions out of a range and periods that never recover. A strategy that realizes many small winners but cannot exit inventory should not be called safe based on its cycle count.

No such paired historical strategy ranking is newly claimed on this page. Our existing grid research is a separate experiment with its own universe and costs; the synthetic comparison here can help formulate a new test, not substitute for it.

Verify: Existing grid profitability research and contract · Pause and reconcile a bot safely

Frequently asked questions

Sources and verification

• 3Commas grid mechanicsOfficial settings, not a profitability endorsement.
• 3Commas DCA mechanicsPrice-triggered purchases and total-order budgeting.
• Transparent synthetic ledgerHypothetical paths, exact fills and cash-paid fee sensitivity.

Explore a bounded AI trading workflow.

Use the paper accounts to inspect recorded behavior; supported historical strategy tests require an account and interpretation review.