Questions for the curious allocator

Long and short positions in USDT-margined perpetual futures on up to 40 of the most liquid cryptocurrencies (11 in the current book) — predominantly Bitcoin, Ethereum, and selected altcoins — at Binance, OKX and Bybit. The programme is directional, not market-neutral: net exposure shifts with the signal composite, and that is how alpha is generated. The active instrument set is reviewed regularly against liquidity, spread and open-interest thresholds; instruments that fail the thresholds are dropped.

Currently 21 as of 30 September 2026, from a set of 10–50 uncorrelated systems approved by the Investment Committee and adjusted to market conditions. They span trend-following, counter-trend, mean reversion, momentum and volatility families, applied across four timeframes (5 minutes, 15 minutes, 1 hour, 4 hours). Capital is allocated on a risk-parity basis so each sub-strategy contributes a similar share of risk rather than being weighted by expected return. The overall capital split was 66% trend-following and 34% counter-trend as at 31 August 2026 — the counter-trend allocation compensates during mean-reverting or choppy periods while trend-following carries the directional book. The Investment Committee resets the split at each quarterly review according to the prevailing market regime, within a range of 50–100% trend-following and 0–50% counter-trend.

Both rules-based and statistical. Most sub-strategies combine rules-based signal constructs with statistical or machine-learned features for parameter selection. All production trading strategies are explicable — opaque black-box models are not deployed without explicit Investment Committee approval, and the IC has not approved any to date. AI agents are used heavily across operations (compliance monitoring, risk reporting, investor support, software development) but not in the trading-signal chain. That describes Algotoria Classic, the live family. The signal of Algotoria Spike, which is pre-production, comes from the firm’s neural-network research; see What is Algotoria Spike?

It thrives in trending and trending-volatile regimes — sustained moves in either direction, with enough volatility for the entry and exit logic to capture range. It struggles in extended low-volatility regimes where signals do not develop, and in chop where directional moves reverse before take-profits trigger. The counter-trend and volatility-filter sub-strategies were added specifically to dampen performance in those conditions. Concrete evidence: during January–February 2026, Bitcoin fell roughly 24% while Algotoria Classic Stable returned approximately +49% gross and Classic Diversified approximately +28% gross — directional short-side moves are favourable conditions, not catastrophic ones.

Every sub-strategy undergoes quarterly review against live-versus-research drift and rolling information-ratio thresholds. Decaying strategies are de-weighted and ultimately retired. New candidates enter through a six-stage validation pipeline: in-sample → out-of-sample → cross-validation → paper-traded incubation → live incubation with size cap → full production. Strategies whose risk-reward collapses in a market shock — cross-exchange arbitrage and similar latency-dependent constructs are the canonical example — are structurally excluded regardless of back-tested Sharpe.

Algotoria Spike is the firm’s next-generation strategy family, built on its neural-network research and deliberately dormant: it opens positions only when model conviction clears a high, frozen threshold, and holds its capital in collateral the rest of the time. Multi-asset, multi-scale models predict direction and volatility, and the signal is pure trend (directional).

It has two strategies, Algotoria Spike Diversified and Algotoria Spike Stable, which take the same trades and differ only in collateral. Both are pre-production: their results are simulated (walk-forward), and neither has managed client money yet.

The two families are positioned differently. Algotoria Spike is the next-generation family: selective, high-conviction and dormant by design. Algotoria Classic is the battle-tested family: a real-money record through different market conditions, and the larger capacity.

Algotoria Spike

• Next-generation
• Selective by design: dormant most of the time, deployed only on high-conviction moves
• Highest expected risk-adjusted performance, not yet fully proven (simulated)
• Best diversification
• Limited capacity (about 2.5 million USD on current estimates)

Algotoria Classic

• Battle-tested through different market conditions, including the January–February 2026 stress window
• Real-money track record since October 2023 (3 years)
• Drawdown budget never breached on the real-money track, gross of fees*
• Larger capacity (about 150 million USD)

* On modelled after-fee bases, Classic Stable’s adjusted maximum drawdown reached −31.3% against its 30% budget.

The side-by-side comparison is on the Strategies page.

Neither Algotoria Spike strategy has managed client money yet. Their records are replays of the Algotoria Spike signal, generated walk-forward, through the production trading engine on exchange candles, so they show how the design would have behaved, not what an investor earned.

A simulation is not a track record: future results will differ, often materially, and all capital committed can be lost. The collateral rule of Algotoria Spike Diversified was chosen on data within the simulation window, so its figures are a hypothesis. Only Algotoria Classic has a real-money track record. Read the Risk Disclosure Notice before allocating.

Rarely, by design. It opens positions only when model conviction clears an entry threshold that was set high in research and then frozen, so it is in the market about 1% of the time, on about 64 traded days a year. Time-weighted, about 99% of its capital sits in collateral. Algotoria Classic, by contrast, is in the market 100% of the time and trades every day.

Algotoria Spike spends most of its time in collateral, and the collateral earns yield, so the simulated figures always include it. Algotoria Spike Stable holds equal thirds of RLUSD, OKUSD and USDG, yield-bearing stablecoins whose rewards accrue daily; the simulation credits them day by day through a collateral ledger, about 4.2% of NAV a year over the window. Algotoria Spike Diversified is credited the same way, plus the gains and losses of its gold and equity sleeve while it is held.

For that reason Algotoria Spike Stable is measured under the NAV-Based Method, so that the stablecoin collateral yield counts in the performance-fee base, since Spike idles most of the time.

The collateral is the only difference, and in the simulation it shows as a shorter and shallower time under water at a similar maximum drawdown and Calmar ratio. Both take the same trades. Spike Stable holds equal thirds of three yield-bearing stablecoins. Spike Diversified holds stablecoins too, and adds gold and a US equity index token under a proprietary allocation rule, selling them again under the same rule.

Neither is the same as the Classic product with the same collateral name. Algotoria Classic Diversified posts a collateral book of stablecoins and Bitcoin; Algotoria Spike Diversified posts yield-bearing stablecoins plus a sleeve of tokenised gold and equity index tokens, and holds no Bitcoin.

Up to 40 USDT-margined perpetual futures; the simulated record trades 7: Bitcoin, Ethereum and five major altcoins. Trading venues are OKX, Binance and Bybit, as for the Classic strategies; the simulation, data and capacity work behind the strategy used OKX. Algotoria Spike trades as a taker, with price-protected marketable orders completed within minutes.

The terms follow the collateral type: Algotoria Spike Diversified takes the terms of Algotoria Classic Diversified, and Algotoria Spike Stable those of Algotoria Classic Stable. The fee schedule, risk tiers, minimum allocations and redemption notice are therefore the same; see the questions on fees, minimum allocation and redemption. Every strategy, Spike included, is delivered through a Separately Managed Account (SMA) or a Managed Sub-Account (MSA).

The PnL method is NAV-Based (AMA §1.1.18(a)) for Algotoria Spike Diversified and Algotoria Spike Stable, as for Algotoria Classic Diversified and Algotoria Classic Stable, so that the collateral yield counts, which matters most for Spike, since it idles most of the time. The Isolated Method (§1.1.18(b)) applies only to Custom-Collateral accounts, Algotoria Spike Custom included.

The same tiers as Algotoria Classic, by risk tier: 15% / 25% / 35% for Algotoria Spike Diversified and 10% / 20% / 30% for Algotoria Spike Stable, with 3× gross exposure at the top tier. This is deliberately conservative for a new strategy.

The budget is a monitored ceiling, not an automatic stop, a guarantee or a contractual floor, and a drawdown can exceed it; the drawdown budget question explains how it is monitored. The simulated record is not an account run at any of these budgets.

Its capacity is limited: about 2.5 million USD on OKX alone on current estimates, before per-order size caps begin to reduce returns; larger accounts need their own assessment. Algotoria Classic has a larger capacity, about 150 million USD.

On the client's own SMA or MSA at Binance, OKX or Bybit, routed by Algotoria's proprietary Execution Engine. The engine prioritises maker-side fills to compound rebates; the current realised mix is about 73/27 maker/taker, with 100% maker execution as a medium-term target, and measured all-in trading cost (fees, slippage and funding) has run at approximately 11 basis points round trip — about half the 20 bps assumed in published simulations. Sub-strategies with similar exposures are concentrated on the same venue where possible, so cross-venue risk does not depend on timely rebalancing.

In a Separately Managed Account (SMA) the client is the master account holder at the exchange and retains full control of API keys; Algotoria is whitelisted with a trade-only key. In a Managed Sub-Account (MSA) Algotoria provisions a dedicated sub-account under its institutional umbrella — the client still owns the assets, but the sub-account usually qualifies for preferential VIP trading fees. For most clients an MSA is the better economics; SMA is the right choice when the client requires direct treasury control over the master account or already holds a VIP tier of their own.

No. The API keys provisioned to Algotoria carry trade-only and read-only permissions. Withdrawal permissions are never requested. Clients create and rotate their own keys for SMAs; for MSAs, Algotoria issues and rotates keys on the client's behalf under the Asset Management Agreement. The separation is structural — no code path in our stack is able to withdraw, transfer or pledge client assets.

Mid-frequency. The average position holds for about 3.6 days. The book trades roughly 50 times per day at portfolio level, and daily traded volume runs at about 1× AUM. We do not run high-frequency market-making or latency arbitrage.

The Execution Engine gracefully degrades. Positions remain in the client's account on the affected venue; orders that cannot be routed are queued until the venue is restored. The firm does not rely on cross-exchange rebalancing as a primary risk control, so a single-venue outage degrades execution but does not generate cascading losses.

Typically one business day from first KYC submission to first live trade. If you fund the exchange account from fiat, add the time the international bank transfer takes to arrive. The sequence is: sign the Asset Management Agreement and acknowledge the Risk Notice → pass KYC / AML → open the exchange account → whitelist Algotoria's API keys → fund the account.

Typical annual drawdowns of 20–25% on Algotoria Classic Stable, 15–30% on Algotoria Classic Diversified. Historical back-tests reached 30%. Drawdowns beyond these ranges trigger a formal Investment Committee review.

All drawdown figures quoted on this site — and the agreed drawdown budget selected during onboarding — are measured on the gross trading-account return curve, before deduction of Algotoria's quarterly performance fee. Net-of-fee drawdowns experienced by the investor are larger by construction. Worked example: for Algotoria Classic Stable over 2024-01-01 → 2026-09-30, the adjusted maximum drawdown is −24.7% gross, −30.2% net of a 25% fee, and −31.3% net of a 30% fee.

Aggregate leverage is capped at 3.0× of account value across all open positions, and is dynamic — it scales down as realised volatility rises. By risk tier, average applied leverage is roughly 100% at High, 67% at Medium, 33% at Conservative; maximum applied leverage is 300 / 200 / 100%. Lower tiers are delivered by scaling leverage down; the strategy logic, signal generation and execution model are otherwise identical across tiers.

The firm monitors 95% one-day parametric VaR at the portfolio level, computed on exchange-matched returns, with a target of approximately 2%. No single sub-strategy may contribute more than 5% of the portfolio's 95% VaR. Realised correlations between sub-strategies are monitored daily; if they drift above Investment Committee thresholds, affected sub-strategies are de-weighted or temporarily suspended. Drawdown thresholds are monitored at the strategy, sub-strategy and instrument level. A breach raises a real-time alert and escalates to the IC; the de-risking itself is then executed by the trading team under the Chief Trader's direction. The only continuously automatic risk reduction is volatility-scaled position sizing and the stop-losses attached to every trade.

The drawdown budget is the target maximum gross drawdown of the account from its most recent high-water mark — the headline number selected during onboarding (10% / 20% / 30% for Classic Stable; 15% / 25% / 35% for Classic Diversified). No drawdown-triggered de-risking is implemented in the trading software and none is applied by default. Drawdown is alerted in real time at 60%, 80% and 100% of the budget; at two-thirds of the budget the Investment Committee reviews the account and decides case by case whether to reduce risk, and at the full budget you are notified and choose whether to continue, reduce the budget or stop trading. The automatic layer is volatility-scaled position sizing and the stop-losses attached to every trade — nothing else de-risks an account on its own. An automated hard stop at a level you specify can be agreed as a custom policy for your account at onboarding. It is an engineered envelope, not a guarantee — the Risk Disclosure Notice publishes the estimated probability of exceeding it. The investor may change the tier at any quarter-end.

Client assets sit at the exchange in the client's own account, so an exchange insolvency directly impacts the client — not the firm. We select venues with proof-of-reserves practices, strong regulation and an established track record (Binance, OKX, Bybit), but exchange exposure remains the client's choice and their risk. For stablecoin risk: Algotoria Classic Stable posts approved stablecoins, including USDT, USDC, RLUSD, OKUSD and USDG, and a full depeg of the stablecoin held would materially impair the account. For stablecoins other than USDT, the account may borrow USDT against them to open USDT-margined perpetual positions, and the borrowed USDT accrues interest at the exchange’s rate, borne by the account; yield-bearing stablecoins typically earn more than this cost. Algotoria Classic Diversified mitigates single-issuer concentration by holding part of the collateral in BTC and tokenised RWAs (e.g. PAXG) — but it does not eliminate it.

Yes. Cryptocurrency derivatives carry the risk of total loss of capital committed. The strategy's risk-parity framework, the staged drawdown-escalation obligations, the dynamic leverage caps and the non-custodial model are designed to keep that probability low — but they do not eliminate it. Black-swan combinations (simultaneous exchange insolvency, stablecoin depeg, prolonged platform outage) can produce losses well beyond the agreed risk ceiling. Do not allocate capital you cannot afford to lose. The full Risk Disclosure Notice is the authoritative document.

TSLab for strategy logic, the proprietary Algotoria Execution Engine for order management, and a service layer on Hetzner and Google Cloud for infrastructure. A 24/7 Network Operations Centre and a Security Operations Centre monitor the stack in real time, with alerts paging the on-call engineer within seconds. The monitoring layer is Zabbix for infrastructure, Prometheus / Grafana for application metrics, and the proprietary Algotoria Monitoring System for strategy-level anomalies.

Your identity documents never leave the EU. KYC files, proof of address and source-of-wealth evidence sit on Algotoria's own servers in a Tier III, ISO/IEC 27001-certified data centre in Germany — segregated from the trading platform entirely and never placed in a public cloud. Trading data sits next to the exchanges. Only what is mechanically needed to place an order — an anonymised account identifier, your encrypted exchange API credentials and the trade records — runs in Google Cloud Tokyo, because the execution engine has to be milliseconds from the Binance, OKX and Bybit matching engines; every extra millisecond of round-trip widens slippage and raises the chance of a failed order. The result is EU-grade handling of your identity and Asia-latency execution on your capital, rather than compromising one for the other.

By construction. Algotoria's stack sends only trade-only API calls. No code path is able to withdraw, transfer or pledge client assets — the API key permissions issued by the exchange do not allow it. Separation is structural, not policy. API keys are encrypted before they ever reach the database, and the master key that unlocks them is held in a separate cloud secrets manager — so database access alone yields nothing usable. Execution keys are additionally IP-whitelisted to Algotoria's execution server, meaning a key copied elsewhere cannot be used at all, and are revoked and re-issued immediately on any suspicion of compromise. Access is strictly role-based and logged.

It means every operational function — software development, risk monitoring, compliance, accounting, investor support and marketing — is executed through AI-assisted workflows under human-in-the-loop oversight. Three concrete outcomes for the investor: the 0% management fee, minutes-not-days response times on investor enquiries and due-diligence requests, and full audit trails preserved in tenant-level logs across every discretionary decision. The Board-approved AI and External Data Provider Policy is aligned with ISO 42001 (AI management systems) and the NIST AI Risk Management Framework. Trading decisions remain rules-based and Investment Committee–approved; AI does not make discretionary trading calls.

No. All production trading signals originate from rules-based or explicable statistical models approved by the Investment Committee. Opaque black-box models are not deployed without explicit IC approval, and the IC has not approved any to date. AI agents support operations — compliance monitoring, risk reporting, investor support, software development — not the trading signal chain itself. That describes Algotoria Classic, the live family. The signal of Algotoria Spike, which is pre-production, comes from the firm’s neural-network research; see What is Algotoria Spike?

The Board approves and reviews the AI and External Data Provider Policy, which governs the use of AI agents across every operational function. Each AI-assisted decision is logged with a full audit trail tied to the tenant and the function. The policy enumerates which models are approved for which workflows, which data classes may and may not be sent to model providers, and the human-in-the-loop checkpoints that gate consequential decisions. Compliance, risk and information-security functions are aligned with ISO 42001 and the NIST AI Risk Management Framework.

Daily time-weighted returns (TWR) with compounding, computed from the unrealised margin balance of each strategy's reference portfolio, denominated in USDT and rebased to 0.00% on 1 January 2024 for the public chart. TWR is the industry-standard methodology that eliminates the distortive effect of capital movements (deposits and withdrawals) on the percentage return. Exchange dashboards (OKX, Binance, Bybit) use simplified estimation methods that do not properly handle transfers, so their headline percentage will differ. The absolute dollar-denominated P&L on the exchange dashboard remains correct; only the percentage is affected.

Via TradeLink Passport, which streams each strategy's reference portfolio through a read-only exchange API. Links to the Classic Diversified and Classic Stable portfolios are on the Classic Diversified and Classic Stable pages. On request, Algotoria will additionally provision read-only API keys to your designated auditor or verification platform, giving direct access to the full trade history and daily return series on the underlying exchange accounts — so you can reconstruct and verify every published metric without relying on firm-generated reports.

Zero management fee, zero entry fee. A quarterly performance fee of 25–30% (tiered by allocation size) on Net Trading Profits above a rolling high-water mark. Formula: P = E − MAX(B, HW) − I, where E is the ending balance, B is the starting balance, HW is the rolling high-water mark, and I is net inflows/outflows. If the account falls below the HWM, no fees are levied until the deficit is fully recovered through trading gains. Clawbacks are explicitly not applied. Classic Diversified and Classic Stable both use the NAV-based variant (AMA §1.1.18(a)); the Isolated USDT variant (§1.1.18(b)) applies only to Custom-Collateral accounts. Fee tiering is documented in Exhibit 1 of the Asset Management Agreement.

By risk tier: $50,000 for High, $100,000 for Medium, $150,000 for Conservative. The same scale applies to both Classic Diversified and Classic Stable. Lower-risk tiers carry a higher minimum because the firm's per-account operating overhead does not scale linearly with leverage; a larger nominal account is required at lower leverage to keep the engagement economically viable. You may change your risk tier at any quarter-end — the strategy logic is identical across tiers, only leverage and the drawdown budget scale. A minimum one-year horizon is recommended; there is no lock-in.

Five business days' notice is required. Partial redemptions are supported. If a redemption falls mid-quarter, a pro-rata performance fee is calculated for the period during which Algotoria actively managed the account, using the same rolling HWM methodology. During severe market stress with large simultaneous withdrawals, the algorithm is programmed to prioritise the liquidation of the most liquid instruments and to scale down portfolio leverage so the outflow is absorbed without generating outsized slippage.

A monthly investor report covering: rolling Calmar ratio, drawdown profile, exposure composition, strategy-level attribution, realised correlations, fees accrued, and net-of-fee return. Quarterly performance-fee statements accompany billing. The Algotoria Telegram bot (@AlgotoriaBot) is the firm's client-facing monitoring interface — on-demand account overview, PnL charts over 24h / 7d / 30d / 90d / 1y / all-time, quarterly reports showing PnL / HWM / fee calculations, multi-account switching for clients with more than one SMA, configurable daily / weekly / monthly notifications, and shareable read-only links for a compliance officer or co-investor. Access is granted per account by your relationship manager. The bot is read-only; it does not execute trades.

Calmar measures return per unit of worst-case drawdown — and the ability to stay allocated through the worst drawdown is the variable that determines realised client outcomes. Sharpe penalises all volatility symmetrically, but upside volatility is not a loss; we report Sharpe too, but Calmar carries the headline. The benchmark set is Bitcoin (BTC-USDT spot, OKX), BITA Crypto 10 (market-cap-weighted top-10 cryptocurrency index, published by BITA GmbH), Gold (PAXG-USDT) and S&P 500 (SPY-USDT-SWAP perpetual on OKX). Bitcoin and BITA C10 represent passive crypto allocations; Gold and S&P 500 give traditional-asset context for investors diversifying across asset classes. Risk-free rate is FRED GS3M.