Marketpawns

Hi everyone!
I'm officially starting a new topic under the actual project name, Marketpawns. (You can find the old discussion thread here )
The project is live at: marketpawns.com
And there are some news:
1. I've recently added a series of four short tutorial videos to make getting started with Marketpawns easier than ever. You can find them all on our youtube channel and also on our educational platform Links: 1 2 3 4
2. We now have x.com account @MarketPawns_com . Please follow us)
3. And I think the most interesting one: we are preparing to launch public monitoring for a couple of our trading strategies. I will announce the specific details in the next few days.

Over the past few days, someone’s bot decided to spam our servers with SQL injections — thousands of requests, hundreds of payloads.
Not a single one got through.
Still, we’re grateful for the “stress test.”
It gave us a reason to recheck our defenses, tighten a few configs, and clean up some leftover files in production.
In a way, every attack is feedback — if you use it right

Just wanted to drop in with a quick status update. Things are buzzing behind the scenes — we're cooking up some new ideas and polish the monitoring service to make the UX smoother.

MarketPawns current list of strategies includes:

1. Trend Following (one is present and one more is in development).
2. Trend Line Breakdown trading.
3. And, the one that’s been our toughest test: Trend Reversal Prediction.

Catching the Reversal — is a high-risk, high-reward move and it requires tricky, complex analysis to filter the signal right that successfully modeling it became a kind of our internal benchmark for proving the power and precision of the algorithms.
So, we believe in testing the hardest thing first, and going to launch the monitoring service starting with this variant.

The strategies we will monitor are still very much experimental. So honestly, it's going to be just as exciting for us as it is for you to watch the data come in!
I'm planning to post some fresh examples of what those reversal signals look like on the charts.

P.S. We're also putting together a newvideo to clearly explain the different trade types and how they are all structured inside the monitoring service.

"Easy to see and understand the leaks (arrows) and emptyings (orange squares) with Ichimoku"

So why do you post it here?

Quick Update & New Video Drop!

Hey everyone! First an update on the youtube channel.
As you know, we recently dealt with a hack attempt. Following that, our new YouTube channel was blocked...
So we’ve fully transitioned back to our older, solid channel. It is now our permanent main channel. All new content (including this video) will be dropping there moving forward. If the new channel is ever restored, we will only use it as a backup/duplicate.
Please subscribe to the main active channel

We released new tutorial on P6rev Trade Type!
This is the trade type used in trading strategies we are going to launch monitoring soon

In this short tutorial

1. What it is P6rev using the External Attraction Model examples.
2. How it works out in 3 Scenarios:
   1. STOP: Hits the Break Level (Immediate Exit).
   2. SAVE: Hits the Bumper Level (Move Stop Loss to zero/plus).
   3. PROFIT: Hits the 1st Target (Take Profit).

We’ve just launched the first version of the Trading Agents Monitoring Dashboard for a selected set of our tactics.

At the moment only a few experimental tactics are online — this is the very first batch.

I’ll publish a short breakdown of the core ideas behind these monitored tactics a bit later, and we’ll keep expanding the list as we go.
A new Monitoring video-tutorial is also coming soon.

Live data: Trading Agents Monitoring

So, a few details about the trading agents that are currently in monitoring and have already executed several trades:
Both agents are driven by the same core strategy — P6rev-type signals toward the 1st Target
During the move toward the target the stop is dynamically adjusted at key events:

1. once a correction of 7% of the model size appears, the initial stop-loss is moved from break level to directly beyond the extreme formed during the bounce from the Reverse Zone
2. once price reaches 15% of the model size, the stop-loss is moved to the 7% level (where 0% = Calculated P6level)

The difference between the agents is in the decision logic at the Bumper Level:

1. Morphy_Pawn calls a neural network to confirm the probability of reaching the 1st Target.
   If the confirmation is negative — the position is closed early (this is the “Taking Early Profit” mechanism explained in this lesson)
2. Tal_First_Move does not request confirmation for the 1st Target — it simply waits either for full target completion or for the tightened stop-loss to be hit.

While the first agents have already started trading, we have also fixed a number of small UI bugs in the Monitoring UI, and expanded the Trading Agents Monitoring Dashboard functionality
- we’ve added search + the ability to switch between card view and table view for agents;
- a small but nice quality-of-life improvement: you can now take screenshots of charts — they will be automatically saved to your Downloads folder.

Right now we’re working on extending the trading agent logic itself — to launch variations of the existing tactics — and preparing two more agents for activation

We’ve just activated two new agents: Capablanca_Starter and Fischer_Novice
They are similar to the previous two — but with one key difference:
instead of fixing the trade at the 1st Target, they trail the stop-loss differently:

1. when price reaches 1st Taret → stop-loss is moved to 15%
2. when price reaches 40% of the model size → stop-loss is moved to 20%
3. the actual fix happens at the 2nd Target = 50%

We just dropped a walk-through on Trading Agents Monitoring.

A quick update
It turned out we accidentally left neural networks disabled in the agent settings — and we only noticed it when we saw a sequence of stop-loss hits this morning
We’ve corrected this setting — and now let’s see how the situation develops from here
On top of that, we’ve also launched two more agents today:
Kasparov_OpenFile
Karpov_Newbie
They are similar to the previous two — but with one key difference:
their first stop-loss adjustment happens after the Bumper Level is reached, not after a 7% correction
Let’s see how this new batch behaves

A quick transparency update on the first weeks of our trading agents’ performance
The initial results turned out catastrophic — with win rates ranging from 11.5% for Capablanca_Starter to 32.3% for Morphy_Pawn.
So we’re not trying to sugarcoat it: this is far below acceptable levels.
Because of that, we’re doing a soft relaunch:

1. all current agents will continue trading on the same strategies,
2. but they will now run on a new neural network,
3. and we’re preparing to launch new agents using a new trading strategy built specifically around this updated NN

We’ll be watching the new setup closely and sharing updates as it unfolds.

Update on our Trading Agents
We’ve just rolled out an important update to the neural network that handles confirmation of the 1st Target.
Going forward, we’re keeping only one agent active for now — Morphy_Pawn — as it already works with the upgraded confirmation model.
Several other agents will be temporarily paused and then relaunched in the coming days using either refreshed models or fully new neural networks designed for better consistency and decision quality.
This update doesn’t affect the core mechanics of the strategy — the P6rev-type signals toward the 1st Target and the dynamic stop-loss logic remain the same.

Coming back with great news and updates soon

We’ve launched a new trading agent on a small live account — here’s its monitoring page: https://marketpawns.com/monitoring/v...s.php?bot_id=1

MarketPawns updates: new UI, Prediction Markets and smarter trade filtering

Hi everyone!))

Over the last few months, our team has been working on a major MarketPawns update. The main goal was to make the platform easier to use, while also making the trading logic more probability-based, risk-aware, and transparent.

We have refreshed the interface and connected the main areas more tightly: model board, analysis, monitoring, community forecasts, wallet, and Prediction Markets now work as one smoother workflow.

One of the biggest additions is Prediction Markets.
This is not designed to replace trading or become a signal-selling gimmick. We see it more as a community layer around the models. When a trading model appears, users can back a forecast using internal Pawn-Coins. For example, users can forecast whether:
- the model reaches its target area,
- price breaks through / overcomes deeper levels,
- the setup reverses or fails after a key level.
The system then tracks the actual model outcome and settles the forecast automatically. So instead of simply saying “I think this setup works”, users can put a small internal stake behind that opinion, build a visible track record, and compare forecast quality over time.
Basic workflow:
1. Pick a model from the board or analysis screen.
2. Open the Prediction Markets / Quick Buy widget.
3. Choose the forecast type you want to back.
4. Commit Pawn-Coins from the wallet.
5. Follow the result in the wallet/history once the model outcome is resolved.
Right now, all newly registered users receive a 100 Pawn-Coin starting bonus, so everyone can try the Prediction Markets system without needing to build a balance first.
We are also thinking about a longer-term idea: developing Pawn-Coin into a real crypto coin/token. This is only at the discussion and research stage for now. We do not want to rush it or make promises before the concept, legal side, tokenomics, and technical model are clear. But the idea is on the table, especially if the internal economy proves useful for community forecasting and model evaluation.

Another major change is in how trading agents decide whether to open a position.
Previously, the logic was much more dependent on simple probability thresholds from our machine-learning models. These models are Keras-based neural-network models trained on historical market/model data. But a probability alone is not enough. A 60% probability can be good or bad depending on how large the potential profit is compared with the potential loss.
So now we use an EV-based filter. EV means Expected Value. In our case, it is basically a coefficient that compares probability-weighted potential profit with probability-weighted potential loss, also taking trading costs such as spread into account.
In simple words: a trade should not only look “likely”. It should also make sense from a risk/reward perspective over a large number of trades. The trading agent now evaluates not only whether the model has a good chance, but whether the expected payoff is good enough compared with the risk.

We also added several neural-network checks at different trade stages.
Instead of making only one decision at the entry point, the system can now re-check an open position at important intermediate levels. These checks are also based on Keras Neural Networks and GBDT machine-learning models. Each model layer is used for a different task: entry filtering, position confirmation, or deciding whether it is better to keep holding or close the trade.
The exact decision logic is internal, but the broad idea is simple: the trade is not treated as “open and hope”. It can be reviewed again when price reaches meaningful market-structure levels.

We have also expanded the monitoring section. It now includes very detailed analytics for the decisions made by the machine-learning models: entry decisions, intermediate checks, decision chains, model responses, weighted profit/loss estimates, and close/keep reasons.
For now, this level of detail is visible because it helps us debug and helps users understand what is happening under the hood. We are not sure yet whether we will keep all of it permanently open in this form, but transparency is an important direction for the project.
Of course, this is still experimental research software and not financial advice. Our goal is not to claim some magic black-box edge. The goal is to make model-based trading more transparent: show the setup, show the forecast, show the risk/reward logic, show the model decisions, and let the history speak over time.

We’ll keep posting updates as the system evolves. Feedback from traders is very welcome!
Especially around how the Prediction Markets layer feels)