2025 02 04 Альберт МК

High-level summary

This was a live trading/debrief session focused on intraday order-flow trading in Russian stocks (Gazprom, Rosneft, Sberbank, VTB, Bashneft, “Pozitiv”/Positive, etc.). Presenter Albert MK and participants analyzed order-book behavior, algorithmic/robot activity, large hidden volumes (iceberg/“Ice”), density levels, and short-term timing patterns to find entry/exit edges.

The emphasis was practical: detect algorithmic patterns (timing, repetition, volumes), trade when liquidity/robots create predictable micro-structures, and manage risk by using partial covers and predefined exits (including “close around zero” when unsure).

Frameworks, playbooks and processes

Order-flow / Algo-detection playbook

• Monitor order book depth for:
  • large visible blocks
  • iceberg orders and hidden size
  • density changes and sudden “take-outs”
• Identify recurring robot timing (example: hits every ~20s, sometimes 40s). Treat the first activation as a trade entry cue.
• Cross-check peers: if a robot turns on in one stock, it often turns on across related names—scan peers.

Entry / exit sizing & risk process

• Enter first when a robot/algo turns on (higher probability to ride momentum).
• Use partial covers: close part of the position quickly when structure proves fragile.
• If uncertain or market action is noisy, close around breakeven.

  “Close around zero” when signals are mixed.

Time-of-day & session tactics

• Watch activation times (examples: ~11:30 start; the last 5 minutes and final minutes before close are special opportunity windows).
• Evening/post-market: robots may continue or shift—check post-market for continuation or news.

Manual verification vs. screener process

• Use screeners but always verify with your own book/flow; screeners may miss large hidden blocks.

Liquidity harvesting / “collecting” process

• Look for “density” consumption (e.g., 30k consumed while 70k remained) to gauge where liquidity will hold or break.
• If a large block is placed (e.g., 700k), monitor closely rather than immediately fighting it.

Key metrics, volumes, timings, and KPIs

• Notable displayed/quoted volumes:
  • 700,000 (participant-level block)
  • 326,000 (Gazprom glass noted post-close)
  • 200,000 and 100,000 blocks referenced multiple times
  • 45,000 (Lao)
  • Robot hits quoted as 12,000 (VTB) and other small repeated prints (e.g., 243, 500)
• Profit / micro-P&L examples:
  • Pulled out ~60 ticks on a Rosneft move.
  • Small intraday wins like “14” ticks, with partial closes as viable gains.
• Timing patterns:
  • Robot hit intervals often around 20 seconds (references also to 19th, 29th, 40th second patterns).
  • Robots may run for 40–90+ minutes and can turn on/off near session phases.
• Market-level readings:
  • Index impact: positive index move caused “Pozitiv” to spike then be sold; index changes can be small (example ~0.07% referenced).
• Operational KPIs:
  • Fill rates vs. displayed depth (how much of a posted block is eaten).
  • Speed/timing to act when robots change cadence.

Concrete examples and tactical takeaways

• If you detect a robot activating (consistent ~20s hits), consider being first to enter—high-probability momentum.
• When a robot or big client order appears in one stock, scan peers—algos often appear across multiple names simultaneously.
• Use order-book density to gauge true interest (example: 30k eaten and 70k remained).
• Don’t fight obvious algorithmic vectors: either join the flow or step back.
• Use the last 5 minutes and final minutes before close as distinct windows for liquidity squeezes/collection points.
• Manual confirmation is critical: screeners sometimes miss hidden size; check live book and tape.
• Take quick small wins (partial take-profits) when the structure proves fragile rather than overholding.
• Watch for “spills” or sudden market-wide dumps—these create opportunities for large, quick topology breaks.
• If a stock shows odd glass behavior (large spread, twitching book), reduce aggression and prefer smaller sizes.

Operational / organizational recommendations (trading desk)

• Develop simple robot-detection rules to flag stocks (e.g., recurring trades every ~20s for >30 minutes; repeated take-outs of specific book levels).
• Create a hand-check checklist to confirm screener alerts: verify block sizes, visible density, time pattern and related names before executing.
• Standardize closing rules for ambiguous market noise (e.g., close around breakeven if signals are mixed).
• Log micro-trades and patterns with timestamps to build a library of repeatable algo behaviors (for future automation).
• Train new traders to read book “twitchiness”, identify iceberg orders and understand index dependency effects.

Limitations and risk notes

• Robots and large players can switch timing and vector; past behavior is not a guarantee.
• Hidden liquidity can mislead; displayed blocks may be only part of the picture.
• Market-wide events (index moves, post-market news) can abruptly negate intraday patterns.
• Behavioral caution: step away if the setup is “not cool” or the order book is “too empty” to move price.

Mentioned examples / case calls

• Rosneft: ~60-tick gain example from entering when the index turned around.
• Gazprom: repeated robot hits (20s, 40s) with glass volumes (326k noted post-close).
• Missed a large crypto spill opportunity—example of the need for continuous attention.
• 700k block presence repeatedly referenced as signaling large participant involvement.

Presenters / sources

• Presenter: Альберт МК (Albert MK)
• Referenced participant: Pavel Bo
• Other discussants: unnamed traders/participants in the live session