| API key | |
| Remaining time-bar queries / day | {{ timebars_counter }} |
| Remaining alt-bar queries / day | {{ altbars_counter }} |
| User plan | Free Basic Pro Unlimited |
| Basic plan | Pro plan | |
| Time-bar
queries / day | 2000 | 5000 |
| Alt-bar
queries / day | 3000 | 10000 |
| Price | 39€ / month | 99€ / month |
Machine learning (ML) algorithms can only achieve their full potential if trained with a sufficient amount of good quality data. Therefore, quality is of utmost importance and that's why in ML there's a popular saying that states: "garbage in: garbage out", i.e. train a model with bad quality data and the predictions will also be of bad quality.
At CryptoDatum.io we make sure that price candle-sticks are error-free by generating them directly from raw trade data. Additionally, our data is continuously challenged with quality tests and the results are published here.
At CryptoDatum.io we offer state-of-the-art alternative bars such as dollar-based or volume-based price bars that display better statistical properties when compared to traditional time-based candlesticks. Learn more about it here.
Time-based candlesticks follow a widespread human-centric data format. Unfortunately, markets do not follow human time rhythms and specially in crypto holds true that in 1 minute can happen as much as in the last 48h.
Altbars are non time-based candlestick bars that allow data aggregation in an efficient manner. This way, your trading bot can have the latest information available when a change in the market occurs, not when the time-based candlestick decides to close.
CryptoDatum.io currently offers data on the most popular cryptocurrency pairs in the Bitfinex exchange, one of the exchanges with the longest trajectory and healthiest volume. Our plan is to keep expanding the amount of offered crypto pairs and exchanges.
All the bar types provided by CryptoDatum.io get updated twice a day, making our datasets ready for back-testing.
For live trading support, please get in contact with us.
Tick bars are generated each time a pre-defined number of transactions takes place. For instance, 1000-tick bars are sampled every time 1000 trades take place in the exchange.
By generating bars depending on the number of transactions instead of time, we are synchronizing our sampling with the flow of information arrival, i.e. if a higher amount of trades takes place within a short period of time, we will sample candles at a higher speed.
The adaptation of the sampling rate according to the number of transactions occurring in the exchange is an extremely useful feature that keeps our trading bot informed of the latest changes in the market.
You can learn more about tick bars here.
Similarly to tick bars, volume bars are generated everytime a pre-defined volume of cryptocurrency is traded in the exchange. For example, in the BTC-USD pair, a 10-volume candle is sampled every time 10 bitcoins are exchanged (which includes both buy are sell orders).
Volume bars address a problem that can arise in tick-bars, which is the arbitrariness of the "tick" definition. Imagine a scenario where hundreds of very small trades take place and consequently lots of tick-candles are sampled. Using volume candles we could solve this issue by only sampling candles when a significant amount of volume of is exchanged.
Dollar bars are sampled every time a pre-defined value is exchanged. This value can be measured in dollars, but it could be also in euros, yens, etc. The value can be calculated as the volume of cryptocurrency exchanged times the price it has been paid for.
Dollar bars apply exceptionally well in very volatile regimes such as in the cryptocurrency markets. For instance, if a cryptocurrency price decreases 50% respect a time point in the past, a person that would like to invest in it could buy twice as much the volume of the cryptocurrency at the current lower price using the same amount of dollars. This exemplifies that using a fix volume bar may be less effective because when the price goes up, less absolute volume may be exchanged due to the more expensive price of the asset. This lower amount of absolute volume would therefore lead to less sampling of volume bars, which would affect the rate at which our bot would be notified of changes in the market.
Instead, if we use dollar bars, we will not be affected so much by the appreciation and depreciation of the cryptocurrency price because we will be measuring the value exchanged instead of the volume, which is somewhat subject to the absolute price of the asset.
Imbalance bars are part of a special type of candles called information-driven bars. The objective of this type of bars is to analyze the sequence of trades/ticks in terms of their sign, price, volume and/or dollar value and trigger a candle sampling every time the sequence and, more specifically, the observed imbalance exceeds our expectations. These expectations are calculated as an exponential moving average (EMA) of sequences of signed ticks observed in previous candles.
The concept of imbalance in this context is used to measure how many trades or ticks are one-sided, meaning how many of these are either buys or sells, or increase/decrease the price. The idea behind this type of bars is that by analyzing the imbalance of the market and how much one-sided it is, we can be notified whenever the imbalance reaches a given threshold so that, theoretically, we can enter on a trading position early enough before a new equilibrium or balance is found.
Imbalance bars come in three flavors: tick (TIB), volume (VIB) and dollar (DIB) imbalance bars. The main difference between them is the nature of the imbalance and expectations, which are expressed as: number of ticks (TIB), specific amount of traded volume (VIB) or specific amount of value/dollar exchanged (DIB).
