Nemeton: Unveiling the Power Behind Our Data
Nemeton: Unveiling the Power Behind Our Data
Intraday traders often lack reliable data to support their decision-making. At frequencies of 5, 10, or 15 minutes, most data sources fail to provide meaningful insights into market movements.
However, certain datasets can offer an edge—one of the most valuable being the order book, which reflects market participants’ positioning and sentiment.
Despite its potential, order book data is typically overlooked by intraday traders due to its complexity. The vast arrays of price and size combinations can be challenging to interpret, and the cost of storage and analysis is prohibitive for most. This is where Nemeton comes in: a processed, flexible, and accessible data suite designed for intraday traders.
By leveraging the expertise of professional quantitative traders, advanced data storage and processing technology, and highly customizable inputs and outputs, Nemeton transforms raw order book data into actionable insights. Whether you’re an institutional or retail trader, Nemeton provides an unparalleled dataset to enhance your trading decisions.
Example: Using Dynamic Order Flow Imbalance for Regime Detection
One of our proprietary trading signals is Dynamic Order Flow Imbalance (DOFI). This metric captures changes in market liquidity by measuring the net increase or decrease in available volume at each order book level. Positive contributions indicate rising liquidity, while negative contributions signal declining liquidity.
DOFI also accounts for price distance from the mid-price:
-
On the bid side, higher prices suggest stronger liquidity.
-
On the ask side, lower prices indicate improved liquidity.
As a dynamic measure, DOFI tracks liquidity shifts relative to the previous observation, effectively identifying changes in buy and sell volumes in dollar terms. This makes it a powerful tool for summarizing how market participants adjust their positions over time.
While its origins lie in high-frequency trading (HFT)—where market makers rely on rapid, evolving signals—DOFI also provides valuable insights for intraday and day traders.
In this example, Order Book Pressure has been sourced daily from Nemeton since February 17, 2024, using the following script.
from aurastream_client.api import get_data
import pandas as pd
key = "API_KEY"
asset = "XRPUSDT_BINANCE_PERPETUAL"
x = get_data(api_key = key, data = {
"ticker_list":[asset],
"start":"2024-02-17 00:00:00",
"end":'2025-03-24 00:00:00',
"interval":"1D",
"source":"nemeton",
"metric":"dynamic-orderflow-imbalance"})
Although not strictly stationary, DOFI tends to fluctuate around zero in normal markets with sufficient liquidity, as shown in the chart below.
A striking observation is the presence of two distinct regimes in DOFI. It is evident that a significant shift occurred in November 2024, introducing a new volatility regime in the indicator. But does this change reflect broader market trends, and can Dynamic Order Flow Imbalance serve as a predictive signal?
To explore this, we will plot the mid-price of the XRPUSDT perpetual from Binance. This data is also available through Nemeton—by simply substituting the metric field with "mid" and specifying 1 level, ensuring that only the best bid and ask prices are considered.
A strong pattern emerges, and the difference is immediately noticeable—DOFI clearly shifts at the exact moment a trader’s eye can detect the change. During strong price moves, market makers adjust liquidity more frequently, reducing attempts to stimulate or fake volume (such as the well-known “spoofing” practice). When volatility is high, there is no need to artificially “stimulate” trades, resulting in distinct liquidity patterns.
But this isn’t the only key insight. We still need to answer the second question: Can we predict such abrupt changes?
The answer is a resounding yes. Evidence suggests that changepoint detection models can effectively identify shifts in distribution, signaling new events in the time series. To achieve this, we use the Pruned Exact Linear Time (PELT) model, a robust method available in the ruptures package.
Without delving too deep into PELT’s technical details, our goal is to detect structural breaks in the time series before they fully establish, ensuring early regime forecasting with confidence.
And the results speak for themselves. The time series is so distinctly divided that it aids the model’s accuracy. But ultimately, our objective is to detect these shifts in real time, allowing us to anticipate new market regimes before they fully unfold.
Below, we provide the corresponding code snippet.
def convert_to_multiindex_dict(data,
exclude_float_variables = [],
drop_lev0 = [],
drop_lev1 = [],
):
# Flatten the dictionary into a list of tuples (asset, time, variable, value)
flattened_data = {}
for asset, times in data.items():
if asset not in drop_lev0:
df = pd.DataFrame(times).T
if drop_lev1:
df = df.drop(drop_lev1,axis=1,errors="ignore")
#Convert to float, unless in exclude
toconv = df.columns.difference(exclude_float_variables)
df[toconv] = df[toconv].astype(float)
df.index = pd.to_datetime(df.index)
flattened_data[asset] = df
df = pd.concat(flattened_data,axis=1)
df.index = pd.to_datetime(df.index)
df.index += add_timedelta
return df
data = convert_to_multiindex_dict(data,
drop_lev1 = ["ObservationTime"],
)
df = pd.DataFrame(data)
y = x.table.droplevel(1,axis=1)[asset]
tgt = y.values
cpd = rpt.Pelt(model="rbf").fit(tgt)
result = cpd.predict(pen=10)
rpt.display(tgt, result)
plt.show()
The chart should look like the below:
What’s particularly interesting is that the red area is identified using a rolling sample, rather than the full dataset, eliminating any look-ahead bias. Despite this constraint, the model still anticipates the shift early enough to provide traders with actionable insights.
Conclusions
The results demonstrate that Dynamic Order Flow Imbalance (DOFI) can effectively signal regime shifts in market liquidity and volatility. By leveraging changepoint detection models like PELT, we can identify structural breaks in real time, without introducing look-ahead bias. This allows traders to anticipate shifts in market conditions early enough to make informed decisions.
Ultimately, combining order flow dynamics with statistical models enhances predictive capabilities, offering a powerful edge in intraday trading.
