--- description: >- Run prediction bots to earn continuously. cover: ../.gitbook/assets/data-farming/predictoordf_main.png coverY: 0 --- # Predictoor Data Farming
**This page** is about Predictoor DF, and [this page](predictoordf-guide.md) is a guide. # About Predictoor DF A “predictoor” runs AI-powered prediction bots in the [Predictoor](../predictoor/README.md) ecosystem. They earn $ from sales of prediction feeds; higher accuracy and higher stake mean higher $. **Predictoor DF** is a substream of Active DF that amplifies predictoors’ earnings, via extra sales to Ocean Predictoor data feeds. Predictoor DF has 37K OCEAN weekly rewards (ongoing) and 100K ROSE rewards (first 4 weeks). The higher baseline sales makes Volume DF and Passive more attractive. ## Introduction **[Ocean Predictoor](../predictoor/README.md)** data feeds predict whether BTC, ETH etc will rise or fall 5min or 1h into the future. These feeds are crowdsourced by “predictoors”: people running AI-powered prediction bots. **[Data Farming (DF)](../data-farming/README.md)** is Ocean’s incentive program, that rewards OCEAN to people who lock OCEAN or do active DF activities. Predictoor Data Farming is a substream of Active DF. It amplifies existing predictoors’ earnings based on their accuracy and stake. ## Predictoor DF Timing Predictoor DF starts counting on Nov 9, 2023, at the beginning of Data Farming Round 62 (DF62). It runs indefinitely. ## Predictoor DF Reward Amounts Predictoor DF has two components: [OCEAN](https://oceanprotocol.com/about-us/ocean-token) rewards and [Oasis ROSE](https://www.coingecko.com/en/coins/oasis-network) rewards. ### OCEAN Rewards OCEAN payouts are 37,000 OCEAN/week, ongoing. - A special “DF buyer” bot purchases Predictoor feeds. It starts operating on Nov 9. Every day, it spends 1/7 of the weekly Predictoor OCEAN budget for another 24h subscription. It spends an equal amount per feed. (Currently there are 20 feeds: 10 x 5min, 10 x 1h.) - The OCEAN comes from the Ocean DF budget, as part of the 75,000 OCEAN/week for Active DF. The Volume DF budget has been adjusted to 37,000 OCEAN/week, and Challenge DF to 1,000 OCEAN/week. Here are details. ### ROSE Rewards ROSE payouts of 100,000 ROSE/week for the first 4 weeks of Predictoor DF. - Payout is at the end of the DF round. Therefore there will be payouts at the end of DF62, DF63, DF64, and DF65. - Payout for a given predictoor is pro-rata to the net earnings of that predictoor over that DF round, specifically (total sales $ to the predictoor) minus (predictoor stake slashed due to being wrong). - The ROSE comes from a generous contribution of[ Oasis Protocol Foundation](https://oasisprotocol.org/) 👪🙏. ## How to Earn $ Via Predictoor DF **Running a predictoor bot will automatically make you eligible for Predictoor DF rewards.** The [Predictoor DF user guide](predictoordf-guide.md) tells how to get started as a predictoor, and how to claim rewards. ## How to Earn More $ Via Passive DF & Volume DF Predictoor DF makes [Active DF](../data-farming/activedf.md) more attractive, and in turn [Passive DF](../data-farming/passivedf.md). Please ensure you're familiar with both before proceeding. **Predictoor DF makes Volume DF more attractive than status quo**, because the volume-based bounds on weekly rewards trends tend to be qualitatively higher for prediction feeds. (The Appendix has details.) Ocean assets with high DCV are easy to identify: it's the 20 OPF-published prediction feeds. This makes the choice of veOCEAN allocation easy: point to those 20 assets. (Only OPF-published feeds are eligible.) **Doing Volume DF gives Passive DF rewards too**, of course. The net result: Predictoor DF means earnings potential from Predictoor DF, Volume DF, and Passive DF. ## Appendix: On DCV Bounds of Prediction Feeds Predictoor DF makes Volume DF more attractive than status quo. Active DF's rewards are bounded by `DCV_bound` which in turn is bound by sales volume and fees. In Predictoor DF, both those factors are raised, which raises the bounding in Volume DF. Let's elaborate. At one time, Volume DF had a "wash consume" problem, where people published and consumed their own datasets. [DF9](https://blog.oceanprotocol.com/data-farming-df8-completed-df9-started-reward-function-tuned-d74b5134b5d1) onwards address this, by putting a bounds on weekly rewards: > `DCV_bound = DCV * m` Where `m` = `DCV_bounding_multiplier` = `Ocean community fee (0.1%) + publish market fee` This stopped wash consume 💪 because it became unprofitable to _do_ wash consume: fees eat up all potential profits. A low `DCV` (data consume volume) or a low `m` (publish market fee) mean low `DCV_bound`. Predictoor DF makes both `DCV` and `m` higher! - **`DCV` is higher:** the 37K OCEAN/week counts as consume volume - **`m` is higher:** publish market fee is 20% for prediction feeds Therefore `DCV_bound` is higher. Specifically: `DCV_bound = 37000 * (0.001 + 0.20) = 7437 OCEAN`. In other words, at least 7437 OCEAN is available for Volume DF in any given week.