The Davanacci sequence is a cyclic, multi-order recurrence relation that generalizes the concepts of the Fibonacci (2-bonacci), Tribonacci (3-bonacci), and Tetranacci (4-bonacci) sequences. Unlike standard n-bonacci sequences, which use a fixed memory for summation, the Davanacci sequence cycles through different summation rules in a periodic three-step rotation.

Developed in March 2026, the sequence is characterized by an oscillating growth ratio produced by the periodic expansion and contraction of its historical look-back window.

The Davanacci sequence ${\displaystyle D_{n}}$ is defined by a 3-step periodic rule applied to previous terms, typically starting from a four-term seed (most commonly ${\displaystyle 0,0,0,1}$). For ${\displaystyle n\geq 4}$, the terms are calculated according to the following phases:

• Step 1 (${\displaystyle n\equiv 1{\pmod {3}}}$): The 2-bonacci Phase (Fibonacci)
  • Rule: Sum the two most recent terms: ${\displaystyle D_{n}=D_{n-1}+D_{n-2}}$.
  • Effect: Provides the "classic" growth curve. It is the most sensitive phase to immediate changes in the sequence.
• Step 2 (${\displaystyle n\equiv 2{\pmod {3}}}$): The 3-bonacci Phase (Tribonacci)
  • Rule: Sum the three most recent terms: ${\displaystyle D_{n}=D_{n-1}+D_{n-2}+D_{n-3}}$.
  • Effect: The "memory" of the sequence expands. Inclusion of older terms causes the growth rate to accelerate compared to the previous step.
• Step 3 (${\displaystyle n\equiv 0{\pmod {3}}}$): The 4-bonacci Phase (Tetranacci)
  • Rule: Sum the four most recent terms: ${\displaystyle D_{n}=D_{n-1}+D_{n-2}+D_{n-3}+D_{n-4}}$.
  • Effect: The sequence reaches its maximum "reach," capturing the broadest range of previous data and typically resulting in a significant numerical jump.

The cycle resets immediately after ${\displaystyle n\equiv 0{\pmod {3}}}$.

Starting with the seed ${\displaystyle (0,0,0,1)}$, the first twenty terms (${\displaystyle D_{0}}$ through ${\displaystyle D_{19}}$) are:

${\displaystyle 0,0,0,1,1,2,4,6,12,24,36,72,144,216,432,864,1296,2592,5184,7776}$

While the Fibonacci sequence converges to the Golden Ratio (${\displaystyle \phi \approx 1.618}$), the Davanacci sequence's ratio (${\displaystyle D_{n}/D_{n-1}}$) oscillates between approximately ${\displaystyle 1.5}$ and ${\displaystyle 2.0}$ based on the current phase.

The sequence achieves high-growth values with a variable computational load. Across a full 3-step cycle, the sequence averages 2 additions per term:

${\displaystyle {\text{Average Additions}}={\frac {(2-1)+(3-1)+(4-1)}{3}}=2}$

In contrast, a standard Tetranacci sequence requires a constant 3 additions per term.

Most sequences maintain a constant growth rate (e.g., the Golden Ratio). The Davanacci sequence utilizes variable acceleration. By cycling through rules, it mimics systems that alternate between conservative resource use and aggressive expansion, making it a model for simulating phenomena like market volatility or population bursts followed by stabilization.

In a standard Fibonacci sequence, a numerical "shock" persists in memory for only two steps; in Tetranacci, it persists for four. Davanacci creates a unique Information Echo: a specific data point influences the next term strongly (2-bonacci), then moderately (3-bonacci), then broadly (4-bonacci), before briefly dropping out of the calculation during the next 2-bonacci phase. This creates a complex rhythm of data retention.

In software engineering, predictable but non-linear growth is utilized for Backoff Algorithms. A "Davanacci Backoff" provides more nuanced "breathing room" for a system to recover (2, 3, 4 units) compared to simple doubling, which can become overly aggressive too quickly.

The cyclic reset adds a layer of complexity for pattern recognition. For generating pseudo-random numbers or "noise," Davanacci is less predictable at a glance than Fibonacci because the underlying summation rule is a moving target, requiring an observer to first identify the rule-change cadence.

By utilizing three different look-back windows, the sequence can model phenomena operating on multiple time scales simultaneously. It bridges the gap between short-term trends (2-term) and medium-term history (4-term). In project management, this can represent how recent obstacles (short-term) and architectural debt (long-term) combinedly influence current velocity.

The sequence was formalized in Huntsville, Alabama by systems engineer David William Stewart II. It was proposed as a framework for Variable Memory Scaling in discrete mathematics and algorithmic design. The 2-3-4 rotation identifies a "compounding momentum" effect where the 4-term summation rule operates on values already accelerated by the preceding 2 and 3-term rules.

The sequence is proposed for use in Network Backoff Algorithms. A "Davanacci Backoff" provides a tiered delay pattern (${\displaystyle 1s,1s,2s,4s\dots }$), allowing for shorter intervals during minor network congestion and longer, high-memory pauses during severe failures.

In cryptography, the sequence is proposed for Memory-Hard Functions. By cycling the look-back window, hardware such as ASICs is forced to re-allocate memory buffers every three steps, increasing the cost of brute-force attacks.

• Generalizations of Fibonacci numbers
• Recurrence relation
• Thundering herd problem